KR20050010099A - Call recommender system for outbound call center - Google Patents

Abstract

PURPOSE: A call recommendation system for outbound call centers is provided to maximize call success rates with clients by estimating an optimized call-available time zone and a channel for each client and attempting a call according to an estimated result. CONSTITUTION: A call recommendation system server(100) downloads a unit calling list from a host computer server(200). The call recommendation system server(100) extracts data about a target client among the unit calling list from a client data DB(130,510). The call recommendation system server(100) determines a call success estimation model. The call recommendation system server(100) grants a call-associated score to each client. The call recommendation system server(100) processes each call-associated score and appoints a call-available time zone for each client. The call recommendation system server(100) transmits the appointed call-available time zone information to a dialing server(300).

Description

Call recommendation system for outbound call centers {CALL RECOMMENDER SYSTEM FOR OUTBOUND CALL CENTER}

The present invention relates to a call recommendation method applicable to an outbound call center, a recording medium recording the method, and a call center system applying the method, and more particularly, a call recommendation method designed to maximize a call success rate with a customer. A recording medium recording a method and a call center system to which the method is applied.

As relationships with customers become an important competitive factor, call centers are now becoming the most important means of active promotion and customer care, as they can only address customers' voices or simply accept and process phone orders. In this trend, the importance of outbound call centers in which call centers call customers is growing, and the size of businesses involved is growing at an enormous rate.

Changes in the IT industry and advances in customer relationship management (CRM) are now moving beyond customer data analysis to bring CRM into a call center, a powerful channel for direct contact with customers. Customer information from agents is stored in ODS (Operational Data Storage) and enables one-to-one marketing through customer information analysis solutions such as data mining. These changes predict that the role of outbound will be increasingly amplified in the future, suggesting that the efficient operation of outbound call centers becomes increasingly important.

Outbound, unlike inbound (the process in which a customer calls the call center), has a low rate of call connection despite the increase in the number of target customers. There is a problem that is not made.

In addition, there is a problem in that the economic capacity to supplement the call center facilities and counselors to meet the rapidly increasing outbound needs also has a problem.

In the current system, where more than 60% of the call center's operating costs are spent as agent management expenses, efficient operation of the counselors is directly related to the reduction of the operating cost of the entire call center. So, an automatic dialing system has been introduced, but there is a need for a more efficient system.

Currently, the outbound call center system sends a list of customers to call (calling list) to the call center every day before the day, and the call center's automatic dialing machine in turn calls the customers and connects them with the agent. In this way, there is a problem that does not consider the characteristics of each customer because they try to call in the order of the customer list, due to this, the call rate is not out of the low state.

Korean Patent Application No. 10-2001-0010541 of Samsung Electronics Co., Ltd. discloses an automatic call distribution event handler for recognizing and processing a call occurrence event, an agent handler for searching whether there is an available agent resource to be allocated to the generated call, and A message handler that allocates the resource to the call if there is an available agent resource, and sends the call to the queue if there is no resource, a queue that stores the call if there is no available agent resource, and periodically An apparatus and method for distributing a call at a call center on a network comprising a timer for searching and controlling a call stored in the queue. By implementing the above invention, the Internet call center service is not limited to the number of gateway channels. However, this is limited to the method and apparatus for distributing calls reached at the call center, and does not include the concept of optimal distribution.

Therefore, we make a call success prediction model for time zone and channel (work phone, home phone, mobile phone, etc.) by using each customer's historical call success history data and personal information, and predict the best available time and channel for each customer. In addition, there is a demand for a system that can maximize the success rate of the call with the customer by attempting to call the customer in the call center according to the predicted result.

The technical problem to be achieved by the present invention relates to a call recommendation method in a call center system applying an optimal call distribution method to increase the call connection rate for customers in the call center.

Another technical problem to be achieved by the present invention relates to a call center system in which the call recommendation method is embedded.

1 is a conceptual diagram of an environment on a network in which a call recommendation system of the present invention operates.

2 is a conceptual diagram of an internal functional unit of the call recommendation system of the present invention.

3 is a flow chart of how the call recommendation system of the present invention operates.

4 is a flowchart illustrating an optimal call time zone designation method according to the present invention.

In order to achieve the technical problem to be achieved by the present invention,

M-1) The call recommendation system server 100 includes a call target and call related information about the call subject to be performed based on a specific unit previously specified from the host computer server 200 storing the call target list. Receiving a call list;

M-2) the call recommendation system server 100 extracts data about a call target customer on the unit call list from the customer data DB 130 or 510;

M-3) the call recommendation system server 100 determining a call success prediction model;

M-4) the call recommendation system server 100 assigns a score related to a call for each customer to the determined call success prediction model;

M-5) the call recommendation system server 100 processes the call-related score for each customer to designate at least one or more of a call time zone and a channel for each customer;

M-6) a call recommendation method in a call center system including the step of transmitting, by the call recommendation system server 100, at least one or more designated information of the call time and channel for each customer to the call originating computer server 300. present.

The customer data DB 130 or 510 of step M-2 may include information on a call center (CTI) history, information on a telemarketing (TM) history, customer basic information, credit information, and an automatic response system of a call target customer ( ARS) may include at least one or more information from the history information, demographic information.

The method of determining the call success prediction model of the M-3) step may include case-based reasoning, decision tree, neural network, spline, and logistic regression analysis. At least one or more of a method including a regression and a cluster analysis may be used.

The method of assigning a score related to a call for each customer in the step M-4) may use a result value obtained by inputting customer information on the customer data DB 130 or 510 into a call success prediction model corresponding function. The method of assigning a score related to a call for each customer may include employing at least one method of scoring by time zone, scoring by communication medium, and scoring by receiving place.

The method of designating a call time zone for each customer by processing the call-related scores for each customer in step M-5) may include inputting at least one score from among time slot scores, scores of communication media, and scores of receiving destinations as input variables. And, as a condition of the maximum number of call customers per time zone, it may be to find the best time-linked customer list by time zone, and the method of finding the time-optimized call-linked customer list by grid time, linear planning method and distance analysis method At least one of the method may be used, and the method of specifying a call time zone for each customer by processing the call-related scores for each customer in step M-5) may include a customer list corresponding to the maximum number of call connection customers per time zone. Extracting to a data structure comprising a currency related score; And sorting the extracted customer list based on at least one or more scores among scores of time slots, scores of communication media, and scores of receiving locations.

The method of designating a call time zone for each customer by processing the call-related score for each customer of step M-5) may be performed using at least one method of a grid search method, a linear programming method, and a distance analysis method.

When assigning a score related to a call for each customer, the method may further include assigning a score including weight information previously assigned for each customer.

The call recommendation method may include: M-7) storing, by the call originating computer, a result of a call originating through the call originating computer;

M-8) may further include the step of reflecting the result of the telephone call in the step M-7) to the customer data DB (130).

(a) having a database storing predetermined score information previously assigned to each of the M time zones when the IDs of the plurality of customers and the day are divided into M time zones of a predetermined number;

(b) if the number of customers is less than or equal to the minimum of the maximum number of customers allowed to connect in each time zone by comparing the minimum value among the maximum number of customers allowed to connect in each time zone, A method of designating an optimal time zone for each customer, comprising: designating a time zone corresponding to a maximum score among the scores of each customer as the optimum time zone of the customer.

(c) if the number of customers is greater than the minimum value of the maximum number of customers allowed to call in each time period, the sum of the maximum number of customers allowed to call in each time period is determined. Comparing the number of customers with each other time period and determining whether the number of customers is less than or equal to the sum of the maximum number of customers allowed for telephone connection;

(d) If the determination result of step (c) is true, when all the CNT values corresponding to the maximum number of customers assigned to each time slot are matched one-to-one with the CNT value assigned to each time slot according to the order, all the CNT values correspond to it. Determining whether the value is less than or equal to a corresponding Max value; And

(e) if the determination result of step (d) is true, the method may include designating a time zone corresponding to the maximum score among the scores of each customer as the optimum time zone of the customer.

(f) if the result of the determination of step (d) is false, determining whether the M value is greater than two; And

(g) if the determination result of step (f) is true,

The M time zones correspond to the predetermined M axis coordinates, the distance values of the M scores of the respective customers are calculated from the axis coordinates of the respective time zones, and the optimum time zones are sequentially assigned from the smallest distance values. And adjusting the optimal time zone given to each customer based on a difference value between the scores of the two optimal time zones of each of the customers assigned to the two optimal time zones, and a rank value that is a rank value of each of the two optimal time zones. can do.

Step (g)

(g1) mapping the M time zones into M axial coordinates as in the following equation;

A_1 = (1, 0, 0, ..., 0), A_2 = (0, 1, 0, ..., 0) ... A_M = (0, 0, ..., 1), where A_1 , A_2, ..., A_M are coordinate axes of each M time zone.

(g2) calculating a distance from a coordinate axis of time zone j for each of the M scores of the customers using the following equation;

Dist _j = {∑ ^M _{k = 1} (Score_k-Kth element of A_j) ² } ^1/2

Where j is any number from 1 to M, M is the number of time zones or scores per customer, and A_j is the axis coordinate of the j-th time zone.

(g3) arranging the obtained distances in ascending order in association with the corresponding customer ID and time zone information;

(g4) generating first optimal time information by assigning optimal time zones in order from the smallest distance, except that the optimal time zone is already given or the maximum allowable number of customers for each time zone is already filled;

(g5) When the value of the arbitrary variable K is changed from 1 to M and the other arbitrary variable L is changed from 1 to M, except that the value of K and the value of L are the same, Extracting, from the first optimal time information, an ID of an adjustment target customer who is K or a customer such as L, a K th score and an L th score of the customer having the ID;

(g6) calculating a difference value and a Rank (R) between two scores of the adjustment target customer; And

(g7) When the number of the adjustment target customers is N, the difference between the scores of the j th customer is D_j, and the rank (R) of the j th customer is R_j, the optimal time zone of the j th customer is the R_j. Is less than or equal to Max_K and D_j is greater than or equal to 0, it may be a method of giving a K-th time zone.

Step (g7)

If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is less than or equal to Max_L, it may be a method of giving the L-th time zone.

Step (g7)

If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is larger than Max_L, it may be a method of giving the K-th time zone.

(g8) The method may further include repeating steps (g5) to (g7) until the optimum time zone of all the customers no longer changes.

Step (g)

(g11) calculating a difference value and a rank value of scores for the two time zones of each customer when the determination result of step (f) is false; And

(g12) N is the total number of customers, and the difference value of the score calculated in step (g1) of the variable j-th customer having an arbitrary value is D_j, and the calculated rank value of the j-th customer is R_j. In this case, the optimal time zone of the j th customer may include designating the first time zone when the R_j value is less than or equal to Max_1 and the D_j value is greater than or equal to 0.

Step (g12) is

When the total number of customers is N, and the difference value of the score calculated in the step (g11) of the variable j-th customer having an arbitrary value is D_j and the calculated rank value of the j-th customer is R_j, The optimal time zone of the j-th customer may be a method of designating a second time zone when the difference value obtained by subtracting the R_j value from the N value is less than or equal to Max_2.

Step (g12) is

When the total number of customers is N, and the difference value of the score calculated in the step (g11) of the variable j-th customer having an arbitrary value is D_j and the calculated rank value of the j-th customer is R_j, The optimal time zone of the j th customer may be designated as the first time zone when the difference value obtained by subtracting the R_j value from the N value is greater than Max_2.

In step (d), if the result of the determination in step (c) is false, the scores of the customers are lowered in descending order until the number of allowed customers in each time slot is filled, except for the customers assigned to any time slot. Increasing the count of the time zone by one while sorting ID and time zone information, and obtaining a first customer set, which is a set of customers assigned to each time zone, and a second customer set not assigned to any time zone; And

In step (e), when all the CNT values correspond to one-to-one correspondence with the CNT values for each time zone and the Max values for each time zone, it is determined whether the CNT values are smaller than or equal to the corresponding Max value. And specifying a time zone corresponding to a maximum score among the scores of each customer belonging to the optimum time zone of the customer.

In step (e), when the CNT values of each time zone and the Max values of the time zone correspond one-to-one, whether all CNT values are smaller than or equal to the corresponding Max value is false. Judge whether or not,

(f) calculating a difference value and a rank value of scores for two time zones of each customer belonging to the first set of customers if the determination result of step (e) is false;

(g) In step (d), the total number of customers belonging to the first customer set is calculated in step (f) of the variable j-th customer having any value of N and customers belonging to the first customer set. When the difference value of the score is D_j and the calculated rank value of the j th customer is R_j, the optimal time zone of the j th customer is equal to or smaller than Max_1 and the D_j value is greater than or equal to 0. If the first time zone is designated;

(h) When the value of the random variable K varies from 1 to M, the customers belonging to the first customer set are extracted the customers with the Kth time zone assigned to the optimal time zone and the customers of the second customer set not assigned the optimal time zone. When the extracted customers are sorted in descending order based on the scores of the extracted K customers' time zones, and the score Score_K rank of the Kth time zones of the j th customers, which is a random value among the sorted customers, is R_j. If the R_j value is less than or equal to the Max_K value, the optimal time zone of the first customer designates the K-th time zone as the optimal time; otherwise, the optimal time zone of the j-th customer is not specified;

(i) When the value of any variable L varies from 1 to M, except for the case where the L value is equal to the K value, among the customers belonging to the first customer set, the L time zone is assigned to the optimal time zone, After extracting the customers of the second unassigned customer set, the extracted customers are sorted in descending order based on the scores of the extracted L-th time zones, and the K-th of the j-th customers is any value of the sorted customers. When the score Score_L rank of the time zone is R_j, the optimal time zone of the j th customer is designated as the L time zone as the optimal time when the R_j value is less than or equal to the Max_L value. Not specifying;

(j) extracting, from the first optimal time information, an ID of the adjustment target customer whose optimal time zone is K or a customer such as L, the Kth score and the Lth score of the customer having the ID, Calculating a difference value and Rank (R) between two scores; And

(k) If the number of customers to be adjusted is N, the difference between the scores of the j th customer is D_j, and the rank (R) of the j th customer is R_j, the optimal time zone of the j th customer is R_j. Is less than or equal to Max_K and D_j is greater than or equal to 0, the method may further include providing a K-th time zone.

Step (k) is

If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is less than or equal to Max_L, it may be a method of giving the L-th time zone.

Step (k) is

If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is larger than Max_L, it may be a method of giving the K-th time zone.

Step (g)

When the total number of customers is N, and the difference value of the score calculated in the step (f) of the variable j-th customer having an arbitrary value is D_j, and the calculated rank value of the j-th customer is R_j, The optimal time zone of the j-th customer may be a method of designating a second time zone when the difference value obtained by subtracting the R_j value from the N value is less than or equal to Max_2.

Step (g)

When the total number of customers is N, and the difference value of the score calculated in the step (f) of the variable j-th customer having an arbitrary value is D_j, and the calculated rank value of the j-th customer is R_j, The optimal time zone of the j th customer may be a method of designating the first time zone when the difference value obtained by subtracting the R_j value from the N value is greater than Max_2.

(l) The method may further include repeating steps (h) to (k) until the optimal time zone of all the customers no longer changes.

Step (e) determines whether all the CNT values are less than or equal to the Max value corresponding to each time when the CNT values for each time zone and the Max values for each time zone correspond one-to-one,

(f) determining whether the value of M is greater than 2 when the determination result of step (e) is true;

(g) if the determination result of step (f) is true,

The M time zones correspond to the predetermined M axis coordinates, and distance values from the M coordinates of the respective time zones are calculated for M scores of the customers belonging to the first customer set, and then the optimal time zones are sequentially selected from the smaller distance values. Is given to each customer belonging to the first customer set based on the difference between the scores of the two optimal time zones and the Rank (R) value of each customer belonging to the first customer set given to any two optimal times. Adjusting the optimal time zone;

(h) When the value of the random variable K varies from 1 to M, the customers belonging to the first customer set are extracted the customers with the Kth time zone assigned to the optimal time zone and the customers of the second customer set not assigned the optimal time zone. When the extracted customers are sorted in descending order based on the scores of the extracted K customers' time zones, and the score Score_K rank of the Kth time zones of the j th customers, which is a random value among the sorted customers, is R_j. If the R_j value is less than or equal to the Max_K value, the optimal time zone of the first customer designates the K-th time zone as the optimal time; otherwise, the optimal time zone of the j-th customer is not specified;

(i) When the value of any variable L varies from 1 to M, except for the case where the L value is equal to the K value, among the customers belonging to the first customer set, the L time zone is assigned to the optimal time zone, After extracting the customers of the second unassigned customer set, the extracted customers are sorted in descending order based on the scores of the extracted L customers' time zones, and the L th of the j th customers, which is any value of the sorted customers. When the score Score_L rank of the time zone is R_j, the optimal time zone of the j th customer is designated as the L time zone as the optimal time when the R_j value is less than or equal to the Max_L value. Not specifying;

(j) extracting, from the first optimal time information, an ID of the adjustment target customer whose optimal time zone is K or a customer such as L, the Kth score and the Lth score of the customer having the ID, Calculating a difference value and Rank (R) between two scores; And

(k) If the number of customers to be adjusted is N, the difference between the scores of the j th customer is D_j, and the rank (R) of the j th customer is R_j, the optimal time zone of the j th customer is R_j. Is less than or equal to Max_K and D_j is greater than or equal to 0, the method may further include providing a K-th time zone.

Step (k) is

If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is less than or equal to Max_L, it may be a method of giving the L-th time zone.

Step (k) is

If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is larger than Max_L, it may be a method of giving the K-th time zone.

(l) The method may further include repeating steps (h) to (k) until the optimal time zone of all the customers no longer changes.

Step (g)

(g1) mapping the M time zones into M axial coordinates as in the following equation;

A_1 = (1, 0, 0, ..., 0), A_2 = (0, 1, 0, ..., 0) ... A_M = (0, 0, ..., 1), where A_1 , A_2, ..., A_M are coordinate axes of each M time zone.

(g2) calculating a distance from a coordinate axis of time zone j for each of the M scores of the customers using the following equation;

[Equation 1]

Dist _{j =} {∑ ^M _{k = 1} (Score_k-Kth element of A_j) ² } ^1/2

Where j is any number from 1 to M, M is the number of time zones or scores per customer, and A_j is the axis coordinate of the j time zone.

(g3) arranging the obtained distances in ascending order in association with the corresponding customer ID and time zone information;

(g4) generating first optimal time information by assigning optimal time zones in order from the smallest distance, except that the optimal time zone is already given or the maximum allowable number of customers for each time zone is already filled;

(g5) When the value of the arbitrary variable K is changed from 1 to M and the other arbitrary variable L is changed from 1 to M, the optimum time zone is excluded, except that the value of K and the value of L are the same. Extracting, from the first optimal time information, an ID of an adjustment target customer who is K or a customer such as L, a K th score and an L th score of the customer having the ID;

(g6) calculating a difference value and a Rank (R) between two scores of the adjustment target customer; And

(g7) When the number of the adjustment target customers is N, the difference between the scores of the j th customer is D_j, and the rank (R) of the j th customer is R_j, the optimal time zone of the j th customer is the R_j. Is less than or equal to Max_K and D_j is greater than or equal to 0, it may be a method of giving a K-th time zone.

Step (g7)

If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is less than or equal to Max_L, it may be a method of giving the L-th time zone.

Step (g7)

If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is larger than Max_L, it may be a method of giving the K-th time zone.

(g8) The method may further include repeating steps (g5) to (g7) until the optimum time zone of all the customers no longer changes.

The method of designating a call time zone for each customer by processing the call-related scores for each customer by the call recommendation system server 100 of step M-5) is to apply the method of any one of the above-described optimal time zone granting methods. Can be.

There may be a recording medium on which a program for executing at least one or more of the above methods is executed by a computer.

There may be a call center system equipped with a program capable of executing at least one of the above methods.

Hereinafter, it demonstrates in detail, referring drawings.

The call center system of the present invention includes a host computer 200 including a call destination list DB; A call recommendation system server 100 including a prediction module 140, a scoring module 150, an optimal distribution module 170, a typical server component 105 including a communication module, and the like; Outgoing computer server 300; The counselor computer 400 and a network 1000 for connecting each computer with the server are included. The call center system of the present invention is shown in FIG.

The call recommendation system 100 transmits the data collection and cleansing module 110, the data mining engine 120, the call recommendation system internal customer data DB 130, the weighted scoring module 160, and the customer list to the call originating computer. The module 180 may further include at least one or more of the transmitting module 180 and the call result report generating module 190.

The data collection and cleansing module 110 may download the customer list from the customer list data DB 210 of the host computer 200, and may collect the personal information and the customer's credit from the customer data DB 510 on the external information server. At least one of the information and demographic information may be downloaded. This is illustrated in detail in FIG. 2. Therefore, the call center system can be connected to an external credit information server, and can extract the credit information of the call target customer from the external credit information server.

In addition, the call recommendation system server 100 of the call center system may include a data mart server therein, and may independently use data results processed by an external data mart server connected to a network.

Hereinafter, the call recommendation method of the present invention will be described with reference to FIG. 3.

Call recommendation method in the call center system of the present invention M-1) call recommendation system server 100 receiving a unit call list from the host computer server 200 that stores the call target list;

M-2) the call recommendation system server 100 extracts data about a call target customer on the unit call list from the customer data DB 130 or 510;

M-3) the call recommendation system server 100 determining a call success prediction model;

M-4) the call recommendation system server 100 assigns a score related to a call for each customer;

M-5) the call recommendation system server 100 processing the call-related scores for each customer to designate a call time slot for each customer;

M-6) the call recommendation system server 100 comprises the step of transmitting the call time designation information for each customer to the call originating computer server 300, each of which will be described in detail below.

In the present invention, the call recommendation system 100 receives a unit call list from the host computer (630). The call recommendation system server 100 accesses the host computer 200 to request data (610), and the host computer 200 extracts the call target list from the customer list data DB 210 (620). It may include.

The unit currency list refers to a currency list including information on the callees to be performed based on a specific unit designated in advance and information on the person. The unit of the list may be a day divided by an integer or other standard such as AM / PM, a week, etc., and the setting of the unit is the currency recommendation system / host computer 200. The operator of may set in advance or the call recommendation system / host computer 200 may set arbitrarily.

The unit call list must include at least one identification information of the call recipient. The identification information may be any unique number assigned to the call recipient, such as a social security number or a member number, a telephone number.

The call recommendation system server 100 extracts data related to the call recipient in the call list received from the host computer 200 from the internal customer data DB 130, and also extracts the data from the customer data DB 510 on the external information server. Can be collected. At this time, the extraction key may be a unique number such as a social security number.

If the customer data DB 130 or 510 is conceptually classified into any data set accessible by the call center system for the customer, 1) information on the call center call history and telemarketing history of the target customer Information, ARS system history, customer basic information, customer credit information, demographic information, and the like.

Customer basic information includes information about access means including telephone numbers that can contact the customer through wired / wireless communication or network, such as the customer's ID, customer name, customer's wired / wireless phone number, work phone number, e-mail address, etc. It may include at least one or more of the personal information, such as the birthday or anniversary of the customer, information about the address or residence.

Information on the call center history of the customer being called includes: past call attempts by time of day, past call success times by time of day, past call objectives achieved by time of day, past call success rate by time of day, past call attempts by channel, and past call by channel. At least one or more of information on past call activity of the call target customer in relation to the call center, such as the number of successes and the inbound call time, may be included. The information and format of the ARS or telemarketing history is similar to the information on the call center history of the customer. The channel is a type of communication network through which a call is mediated, for example, a landline phone or a wireless phone. For example, a point of contact with a customer may be an example of a channel such as a home phone number or a work phone number. .

The basic customer information may include at least one piece of information about the user's personality such as age, residence code, gender, and occupation.

The customer credit information may include at least one or more pieces of information related to the customer's financial life, such as credit rating, various credit scores, credit card limit, loan limit, past transaction type, past delinquent amount, past delinquency, and holding card type. Can be.

Demographic information refers to statistically processed data on pre-investigated populations for specific groups or entire populations. For example, women in their 50s or older are 20% more likely to be home in the morning than those in the afternoon, or that college students from their 20s are statistically less likely to have early weather. Such demographic information may be obtained from computer servers of the National Statistical Office or credit information companies, credit card companies or marketing companies.

The call recommendation system 100 combines call center history data for each customer, basic information for each customer, and credit information for each customer to create modeling data by a table join method. The outliers and missing values are then processed to correct the incorrect data. The data is then transformed or derivatives are generated to increase prediction accuracy. The above process may be performed by the data collection and cleaning module 110.

We then apply data mining algorithms, utilize the most predictive models, or combine multiple models to determine the score for each channel for each customer. The role as described above may be performed by the data mining engine 120 (650).

In this manner, the call recommendation system 100 determines a call success prediction model based on the data of the call target customers on the unit call list (660). In this prediction model, the objective variable may be call success in each time zone and / or call success in each channel. Hereinafter, a prediction model that can be used in the present invention will be described.

First, it is a case-based reasoning model using only historical currency success rate data. This model uses an algorithm that predicts the time and channel of the customer's optimal time and channel for the time and channel that the customer has attempted to call for the past period.

Second is the decision tree model. Decision trees are a technique mainly used for classification of data mining, and analyze the data collected in the past to classify patterns existing between them, predict the value of the class, and express the classification form in tree shape. . Decision trees have a disadvantage in that the shape of the model varies with different data, but it is easy to understand because it gives the basis for classification or prediction, and it is easy to understand which attributes have a decisive effect on each class value. It is a popular technique because of its advantages. For personal call recommendation, the predictive model using decision tree algorithm, which is widely used, is used as the objective variable of the success of call of past data as the explanatory variable. Ensemble algorithms such as Bagging, Boosting, and Arcing can be used to increase the accuracy of the prediction.

Third is the logistic regression model. Logistic regression is a widely used statistical method for estimating the function of binary objective variables. For personal call recommendation, we use the predictive model applying the logistic regression analysis algorithm as the objective variable as the success variable of past data as the objective variable and the customer-defined information as the explanatory variable.

Fourth, neural network model. The neural network model is a model that mimics the nerve cells of the human brain, modeling a network structure consisting of nodes and rings, and finding patterns inherent in data through iterative learning process from the data collected in the past like decision trees. Technique. Neural networks are data mining techniques widely used in tasks such as classification, clustering, and association rule discovery. They are applied in various fields such as credit evaluation, card theft pattern analysis, demand and sales forecasting, and customer segmentation. For personal call recommendation, the predictive model using neural network algorithm is used with the success variable of past data as the objective variable and the customer-defined information as the explanatory variable.

Fifth, it is a spline model. Spline is an accurate function estimation method that fits nonlinear functions for objective variables. For individual call recommendation, the Spline prediction model using Dichotomous Regression algorithm is used, with the success variable of the past data as the objective variable and the previously defined customer information as the explanatory variable.

Sixth, cluster analysis model. Cluster analysis is a method of dividing a population into groups with similar properties. That is, it is an analysis technique that maps similar data to the same group. Cluster analysis simply separates the data from each other, and does not fit the classification result to the known data, but uses the classification result to separate unknown data. Algorithms used in this method are often used for self-organizing feature map (SOM) and k-means algorithm. For the recommendation of individual call, we use cluster analysis result using similarity of customer information.

The call recommendation system 100 may determine the prediction model 141 for the optimal time zone using at least one or more of the call success prediction models to score the optimal time zone for each customer (scoring, scoring, scoring). (680). In this case, the scoring may be performed by determining a prediction model for each optimal channel, and the scoring may be performed by combining a time zone and a channel.

When the call recommendation system 100 assigns a score related to a call for each customer, the call recommendation system 100 may assign a score including weight information previously assigned to each customer (720). In other words, a particular customer or group of customers may have a different weight of currency than other customers. Factors for determining the weight may be based on the importance of the customer, the urgency of the call, or weight information input or designated by a person using the call recommendation system or the call recommendation system.

In this process, the predictive model 142 of the possibility of achieving the telephony purpose may be utilized. That is, a prediction model corresponding to each element for determining the weight and a telephony purpose may be applied and applied to the scoring process in consideration of the weight.

Such weight information may be incorporated into the scoring to give the final score. This process may be used only selectively (700).

The call recommendation system 100 may assign a score based on at least one of each time slot and channel related to a call for each customer. For example, as follows.

An example of integrated customer data Social Security Number gender age Past morning call rate Past afternoon call rate . . . One 680213-xxxxxxx One 35 0.33 0.60 . . . 2 451123-xxxxxxx 2 58 0.20 0.50 . . . 3 700604-xxxxxxx 2 33 0.75 0.10 . . . 4 741012-xxxxxxx One 29 0.60 0.66 . . .

Scores can be calculated by integrating integrated customer data, such as Table 1 above, into a predictive model. The predictive model may aim to calculate scores over time and / or per channel. One example of the predicted result is in Table 2.

Example of Calculating Scores by Time and Channel Using Data Mining Models Social Security Number AM score PM score Home score Mobile score . . . One 680213-xxxxxxx 0.67 0.47 0.28 0.63 . . . 2 451123-xxxxxxx 0.35 0.21 0.72 0.57 . . . 3 700604-xxxxxxx 0.76 0.47 0.36 0.44 . . . 4 741012-xxxxxxx 0.29 0.66 0.32 0.59 . . .

The call recommendation system 100 may process a call-related score for each customer to designate a call time slot for each customer, which will be described in detail below. The reason why the call time zone is specified is as follows. Suppose that the maximum call volume in a half-day call center is two. As shown in Table 2 above, when it is advantageous to call all mornings 1, 2, and 3, the question arises: who and whom to choose and in what order is most efficient. In other words, given the limitations of the call center (here, the maximum call volume of a half-day call center 2), it is a question of which time and which channel is most effective for individual call customers. Therefore, determining this optimal distribution problem is to designate the optimal call window for each customer.

A module that performs an optimal distribution process is called an optimal distribution module 170. The optimal distribution module can perform optimal distribution only for the time zone. Furthermore, the optimal distribution for each channel can be applied, and the optimal distribution can be made by reflecting the time zone and the channel in combination.

Hereinafter, the optimal distribution will be described based on the process of designating the optimum call time for each customer.

The method of designating an optimal call time slot for each customer borrowed by the present invention may be classified into a grid search method, a linear program method, and a distance method.

The grid search method selects the combination whose maximum objective function is calculated by calculating the objective function values for all possible combinations satisfying a given condition.

Linear programming is a method of maximizing the objective function under the constraints of first-order inequalities, such as whether it can be effectively and appropriately distributed and problems with production planning.

Distance analysis is a method of maximizing the objective function through the distance between scores. When the size of the analysis target is large, the objective function is selected in consideration of the processing speed and accuracy.

Since the grid search method and the linear programming method have a disadvantage in that the processing speed increases exponentially as the number of customers increases, the present invention employs a distance analysis method when the number of customers is large.

Hereinafter, the flow chart shown in FIG. 4 will be described.

First we define the term. M: Number of time zone divisions, Max_M is the maximum number of customers allowed in M time zone, CNT_M is the current number of customers assigned in M time zone, min () is the minimum number of allowed customers in all time zones, and sum () is the allowed customer in all time zones. The sum of the numbers, R_j, is the ranking of the scores of customers in a particular time zone.

The optimal time zone designation requires a database storing predetermined score information previously assigned to each of the M time zones when the IDs of a plurality of customers and the day are divided into a predetermined number of M time zones. This preparation step is called step (a).

In addition, the maximum number of customers allowed to connect in each time zone must be specified in advance. This maximum number of customers can be an important limitation.

If the number of customers is less than or equal to the minimum value of the maximum number of customers allowed to call each time zone by comparing the minimum value among the maximum number of customers allowed to call each time zone (this determination step) It is referred to as the determination of step (b)), Method 1 of the detailed determination of the present invention can be applied. Algorithm 1 of the present invention is characterized in that a time zone corresponding to the maximum score among the scores of each customer is designated as the optimum time zone of the customer.

If the number of customers is greater than the minimum value of the maximum number of customers allowed to call each time zone, the sum of the number of customers and the maximum number of customers allowed to call each time zone is determined. In comparison, it is necessary to determine whether the number of customers is less than or equal to the sum of the maximum number of customers allowed for telephone connection for each time period. This judgment step is called judgment of step (c).

If the determination result of step (c) is true, all CNTs when one-to-one correspondence of the CNT value, which is the number of customers allocated to each time zone, in one-to-one correspondence with the maximum number of allowed customers for each time slot, in that order, using the maximum score. It is necessary to determine whether the value is less than or equal to the corresponding Max value. (This judgment step is called judgment of step (d).)

If the determination result of step (d) is true, Algorithm 1 is applied.

If the determination result of step (d) is false, it is necessary to determine whether the M value is greater than two. (This judgment step is called judgment of step (f).)

If the determination result of step (f) is true, algorithm 3 is applied, otherwise algorithm 2 is applied.

Algorithm 3 is applied when the time zone is 3 or more, and has the following characteristics. Algorithm 3 goes through the initial value generation step and the fine tuning step. First, an initial value generation step will be described, and this step is composed of the following steps (g1) to (g8).

The M time zones are mapped to M axial coordinates as in the following equation. (This step is referred to as step (g1).);

A_1 = (1, 0, 0, ..., 0), A_2 = (0, 1, 0, ..., 0) ... A_M = (0, 0, ..., 1), where A_1 , A_2, ..., A_M are coordinate axes of each M time zone.

The distance Dist _j with respect to M coordinates of each customer is calculated using the following equation (this step is referred to as step (g2));

[Equation 1]

Dist _{j =} {∑ ^M _{k = 1} (Score_k-Kth element of A_j) ² } ^1/2

Where j is any number from 1 to M, M is the number of time zones or scores per customer, and A_j is the axis coordinate of the j time zone.

Sorting the obtained distances in ascending order in association with the corresponding customer ID and time zone information (this step is referred to as step (g3));

Except when the optimal time zone has already been given or the maximum number of allowable customers for each time zone has already been filled, the first optimal time information is generated by sequentially assigning the optimum time zones from the smallest distance (this step (g4)). It is called.)

Then, the process proceeds to the adjustment step, which can be expressed by the following algorithm.

Do K = 1 to M

Do L = 1 to M (where K is different from L)

(1) ID, Score_K, Score_L extraction only for customers with optimal time zone K or L

(2) Calculate the difference between two time zone scores for each customer (D = Score_K-Score_L) and Rank (R)

(3) If the number of extracted customers is N, the difference between the score of j th customer is D_j, Rank is R_j

(3-1) If (R_j <= Max_K and D_j> = 0) Then (best time zone of jth customer = K)

(3-2) Else If (N-R_j + 1 <= Max_L) Then (best time zone of jth customer = L)

(3-3) Else If (N-R_j + 1> Max_L) Then (best time zone of jth customer = K)

End

End

Then, repeat the above process until the optimal time zone for all customers no longer changes.

The adjustment stage is described as follows.

When the value of any variable K is changed from 1 to M, and when any other variable L is changed from 1 to M, except that the value of K and the value of L are equal, the optimum time zone is K or Extracting the ID of the adjustment target customer that is the customer, such as L, the K-th score and the L-th score of the customer having the ID from the first optimal time information. (This step is called step (g5).)

It is a step of calculating the difference value and Rank (R) between two scores of the adjustment target customer. (This step is called (g6) step.)

If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is greater than the Max_K. If less than or equal to and D_j is greater than or equal to 0, it is given to the K-th time zone, and if the value of N-R_j +1 is less than or equal to Max_L, it is given to the L-th time zone, and if the value of N-R_j +1 is greater than Max_L It is a step to assign to the K-th time zone. (This step is called step (g7).)

Repeating steps (g5) to (g7) is performed until the optimal time zone of all the customers no longer changes (this step is referred to as step (g8).)

Algorithm 2 is applied when the time zone M is 2. Algorithm 2 can be expressed as follows.

Calculate the difference (D = score_1-score_2) and rank for each customer's two time zones, letting the total number of customers be N, the jth customer's score difference D_j, and rank R_j Give an optimal time zone as follows:

(1) If (R_j <= Max_1 and D_j> = 0) Then (best time zone of jth customer = 1)

(2) Else If (N-R_j + 1 <= Max_2) Then (best time zone of jth customer = 2)

(3) Else If (N-R_j + 1> Max_2) Then (best time zone of jth customer = 1)

The algorithm 2 will be described as follows.

When the determination result of step (f) is false, when M is 2, the difference value and the rank value of scores for the two time zones of each customer are calculated. g11) referred to as step.)

When the total number of customers is N, and the difference value of the score calculated in the step (g11) of the variable j-th customer having an arbitrary value is D_j and the calculated rank value of the j-th customer is R_j, The optimal time zone of the j-th customer is designated as the first time zone when the R_j value is less than or equal to Max_1 and the D_j value is greater than or equal to 0, and the difference value obtained by subtracting the R_j value from the N value is less than or equal to Max_2. Is designated as the second time zone, and if the difference value obtained by subtracting the R_j value from the N value is greater than Max_2, the first time zone is designated (this step is referred to as step (g12)).

If the determination of step (c) is false, that is, if the number of customers is greater than the sum of the maximum number of customers allowed to connect to each time zone, the following algorithm 4 is applied.

Algorithm 4 is as follows. Except for customers assigned to a certain time zone, counting the time (count, CNT) of the time zone while sorting the ID and time zone information of the customer in descending order until the number of allowed customers in each time zone is filled is 1 Step by step to obtain a first customer set, which is a set of customers assigned to each of the time zones, and a second customer set not assigned to any time zone (this step is referred to as step (e)).

In step (e), if the CNT values of each time zone and the Max time values of the time zone correspond one-to-one, it is determined whether all the CNT values are smaller than or equal to the corresponding Max value, and if it is true, algorithm 1 is applied to the first set of customers. In this case, a time zone corresponding to the maximum score among the scores of each customer belonging to the first customer set is designated as the optimum time zone of the customer.

When one-to-one correspondence between CNT values of each time zone and Max time zones is determined one-to-one, it is determined whether all CNT values are smaller than or equal to the corresponding Max value, and if it is false, it is determined whether the value of M is greater than two. If M is not greater than 2, algorithms 2 and 5 are applied in combination. Algorithm 5 refers to steps (h) to (k) described below. This is described as follows.

If M is not greater than 2, the following process is performed.

Calculating a difference value and a rank value of scores for two time zones of each customer belonging to the first set of customers;

In step (e), scores for two time zones of each customer are set for a variable jth customer having an arbitrary value of N and customers belonging to the first customer set as the total number of customers belonging to the first customer set. When the difference value of D_j and the calculated rank value of the j-th customer is R_j, the optimal time zone of the j-th customer is less than or equal to Max_1 and D_j is greater than or equal to 0. If the difference value obtained by subtracting the R_j value from the N value is less than or equal to Max_2, the second time zone is specified. If the difference value obtained by subtracting the R_j value from the N value is greater than Max_2, the first time zone is specified. Specify the time zone.

Then, when the value of the random variable K varies from 1 to M, after extracting the customers who belong to the first customer set, the Kth time zone is assigned to the optimal time zone, and the customers of the second customer set that are not assigned the optimal time zone, The extracted customers are sorted in descending order based on the scores of the extracted customers in the Kth time zone, and the jth score when the score Score_K rank of the jth customer, which is any value among the sorted customers, is R_j. If the R_j value is less than or equal to the Max_K value, the customer's optimal time zone specifies the K-th time zone as the optimal time. Otherwise, the customer's optimal time zone does not specify the j-time customer's optimal time zone (this is referred to as step (h)). .)

When the value of any variable L varies from 1 to M, except for the case where the L value is equal to the K value, the customer whose L time zone is assigned as the optimal time zone and the optimal time zone are not assigned among the customers belonging to the first customer set After extracting the customers of the second customer set, the extracted customers are sorted in descending order based on the scores of the L-th time zones of the extracted customers, and the score of the k-th time zone of the j-th customer, which is any value of the sorted customers. When the score_L rank is R_j, the optimal time zone of the j th customer is designated as the optimal time zone when the R_j value is less than or equal to the Max_L value. Otherwise, the optimal time zone of the j th customer is not specified. (This is called step (i).)

The ID of the adjustment target customer whose optimal time zone is K or the same as L, the Kth score and the Lth score of the customer having the ID are extracted from the first optimal time information, and between the two scores of the adjustment target customer. Calculate the difference value and Rank (R). (This is called step (j).)

If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is greater than the Max_K. If less than or equal to and D_j is greater than or equal to 0, it is given to the K-th time zone, and if the value of N-R_j +1 is less than or equal to Max_L, it is given to the L-th time zone, and if the value of N-R_j +1 is greater than Max_L The step of giving a K-th time zone is further included. (This is called step (k).)

Then, steps (h) to (k) may be repeated until the optimal time zone of the customers does not change.

In Algorithm 4 or lower, when one-to-one correspondence between CNT values of each time zone and Max time values of each time zone is determined to determine whether all CNT values are smaller than or equal to the corresponding Max value, and false, and M is greater than 2, 5 is applied in combination. It demonstrates in detail below.

The M time zones correspond to the predetermined M axis coordinates, and distance values from the axis coordinates of each time zone are obtained for M scores of the customers belonging to the first customer set. And a value assigned to each customer belonging to the first customer set based on a difference value and a Rank (R) value between the scores of the two optimum time bands of each customer belonging to the first customer set given to any two optimal times. There is a step to adjust the optimal time zone. This step corresponds to algorithm 3.

When the value of the random variable K varies from 1 to M, after extracting the customers of the first customer set whose Kth time zone is assigned as the optimal time zone and the customers of the second customer set that are not assigned the optimal time zone, the extraction is performed. When the extracted customers are sorted in descending order based on the scores of the K-th time zones of the registered customers, and the score Score_K rank of the k-th time zone of the j-th customer, which is any value among the sorted customers, is R_j, The optimal time zone may specify the K-th time zone as the optimal time when the R_j value is less than or equal to the Max_K value, and otherwise, does not specify the optimal time zone of the j-th customer. This step is the same as step (h) above.

When the value of any variable L varies from 1 to M, except for the case where the L value is equal to the K value, the customer whose L time zone is assigned as the optimal time zone and the optimal time zone are not assigned among the customers belonging to the first customer set After extracting the customers of the second customer set, the extracted customers are sorted in descending order based on the scores of the L-th time zones of the extracted customers, and the score of the k-th time zone of the j-th customer, which is any value of the sorted customers. When the score_L rank is R_j, the optimal time zone of the j th customer is designated as the optimal time zone when the R_j value is less than or equal to the Max_L value. Otherwise, the optimal time zone of the j th customer is not specified. There are steps that do not. This step is the same as step (i) above.

The ID of the adjustment target customer whose optimal time zone is K or the same as L, the Kth score and the Lth score of the customer having the ID are extracted from the first optimal time information, and between the two scores of the adjustment target customer. There is a step of calculating the difference value and Rank (R). This step is the same as step (j) above.

If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is greater than the Max_K. If less than or equal to and D_j is greater than or equal to 0, it is given to the K-th time zone, and if the value of N-R_j +1 is less than or equal to Max_L, it is given to the L-th time zone, and if the value of N-R_j +1 is greater than Max_L It is given as the K time zone. This step is the same as step (k) above.

Steps (h) to (k) may be repeated until the optimal time zone of the customers no longer changes.

In the above description when all CNT values are less than or equal to the corresponding Max value when the CNT values for each time zone and the Max time values for each time zone are one-to-one correspondent to false, and M is larger than 2, Algorithm 3 Can be applied in conjunction with Algorithm 5.

When the optimal distribution module performs the optimal distribution process, the optimal distribution of the call object is determined (740).

When the optimal call time zone designation process for each customer is completed, the call recommendation system server 100 transmits the call time zone designation information for each customer to the call originating computer server 300 (750).

The call originating computer server 300 The call originating computer server 300 proceeds to the customer based on the specified call target information, and then connects the call to the individual counselor computer 400 (760).

The call originating computer server 300 or the individual counselor computer 400 may store call results alone or in combination. In this process, the call originating computer server 300 may receive the call result automatically or the individual counselor computer 400 may be manually input by the individual counselor. The call result may be transmitted to the call recommendation system 100, and the transmitted result may be stored in the customer data DB 130. The call recommendation system 100 may update the data on the prediction module, the optimal distribution module, and the scoring module 150 by reflecting the call result.

The call recommendation method has been described based on the optimum call time of the customer. If you use the above best call recommendation method, you can find the best call channel. This can be achieved by placing the optimal channel in correspondence with the call window at the level of those skilled in the art. The channel refers to a kind of communication means accessible to a customer including a residential telephone number such as a home telephone number, a workplace telephone number such as a company, a mobile telephone number or other voice communication number, and a video communication number.

Such channels may also include non-real time communication means with customers such as e-mail, fax, short messages, one-way or two-way messages. This may naturally include a communication means such as a messenger. In addition, such a channel may be utilized according to a communication means system according to the location of the customer. For example, a distinction may be made between domestic or international calls, local or long distance calls, and the like.

The reason why such a channel is meaningful is that the selection of the optimal communication means according to the communication cost according to each communication means and the characteristics of each customer (physical location, age, type of communication means and whether or not it is held) is important in call center operation. This may be particularly useful when a real-time two-way communication means such as a telephone and a non-real time communication means such as an electronic mail or a non- two-way communication means such as a short message are combined.

Furthermore, the call recommendation method may also be applied to a method that considers an optimal time zone and an optimal channel together. That is, the target channel may be allocated to a more detailed time zone by targeting the internal customer list having the optimal time zone determined, or the optimal time zone may be allocated to the customer list to which the optimal channel is assigned.

All of the above methods can be recorded in a specific recording medium by a computer program. Such a recording medium may be storage or recording means that can be used in a conventional computer system such as HDD, CD-ROM, DVD, MEMORY STICK, RAM, ROM, and the like.

Furthermore, the call center system itself including such a recording medium may be included in the scope of the present invention. That is, the call center system of the present invention may include a program therein capable of executing all the above optimal call distribution methods. The term “inside” is not necessarily physically internal. In other words, even if there is a recording medium that executes some of the above methods on the third computer that the call center system is connected to the network or contains a program capable of executing the above methods, it is conceptually referred to as the interior of the call center system. Because it can.

Example

Examples are described for the understanding of the present invention. However, it is obvious that the present invention is not limited to this embodiment.

Application of Algorithm 1.

According to Algorithm 1, the time zone corresponding to the maximum score of each customer's score is designated as the customer's optimal time zone.

Score for each time zone for each customer (algorithm 1 applied) ID score_1 score_2 . . . score_M 6305241032934 0.723 0.394 . . . 0.429 4611052130294 0.596 0.692 . . . 0.265 7204121064591 0.339 0.218 . . . 0.506 5812042103428 0.618 0.502 . . . 0.715 6605122130294 0.593 0.215 . . . 0.443 7108222014391 0.448 0.733 . . . 0.504

Best time zone for each customer (algorithm 1 applied) ID Best time zone 630524 One 461105 2 710822 M 630524 M 630524 One 630524 2

Application of Algorithm 2

This is the result of applying Algorithm 2 when M = 2, Max_1 = 2, and Max_2 = 5.

Scores for each time zone for each customer (algorithm 2 applied) ID Score_1 Score_2 6305241032934 0.723 0.394 4611052130294 0.596 0.692 7204121064591 0.339 0.218 5812042103428 0.618 0.502 6605122130294 0.593 0.215 7108222014391 0.448 0.733

Difference and rank of score for each customer (algorithm 2 applied) ID Difference in score Rank 6305241032934 0.329 2 4611052130294 -0.096 5 7204121064591 0.121 3 5812042103428 0.116 4 6605122130294 0.378 One 7108222014391 -0.285 6

Best time zone for each customer (algorithm 2 applied) ID Best time zone 6305241032934 One 4611052130294 2 7204121064591 2 5812042103428 2 6605122130294 One 7108222014391 2

Application of Algorithm 3

Application result for the case of m = 3 or more.

Score for each customer for each time zone (algorithm 3 applied) ID Score_1 Score_2 . . . Score_M 6305241032934 0.723 0.394 . . . 0.429 4611052130294 0.596 0.692 . . . 0.265 7204121064591 0.339 0.218 . . . 0.506 5812042103428 0.618 0.502 . . . 0.715 6605122130294 0.593 0.215 . . . 0.443 7108222014391 0.448 0.733 . . . 0.504

Distance for each customer (algorithm 3 applied) ID Street_1 Distance_2 . . . Distance_M 6305241032934 0.762 1.214 . . . 0.992 4611052130294 1.526 0.989 . . . 2.235 7204121064591 1.734 1.982 . . . 1.021 5812042103428 0.973 1.718 . . . 0.879 6605122130294 1.020 1.922 . . . 1.769 7108222014391 1.002 0.814 . . . 0.993

Initial time zone for each customer (algorithm 3 applied) ID Initial time zone 6305241032934 One 4611052130294 2 7204121064591 M 5812042103428 M 6605122130294 One 7108222014391 2

Best time zone for each customer (algorithm 3 applied) Street ID Best time zone One 0.762 6305241032934 One 2 0.814 7108222014391 2 3 0.879 5812042103428 M 4 0.973 5812042103428 One 5 0.989 4611052130294 2 6 0.992 6305241032934 M 7 0.993 7108222014391 M 8 1.002 7108222014391 One 9 1.020 6605122130294 One 10 1.021 7204121064591 M

Application of Algorithm 4

Applies when the number of call list customers is greater than Max_1 sum of Max_M.

Scores for each time zone for each customer (algorithm 4 applied) ID Score_1 Score_2 Score_3 6305241032934 0.723 0.394 0.429 4611052130294 0.596 0.692 0.265 7204121064591 0.339 0.218 0.506 5812042103428 0.618 0.502 0.715 6605122130294 0.593 0.215 0.443 7402172104936 0.493 0.619 0.810 7108222014391 0.448 0.733 0.504

Time zone for each customer (algorithm 4 applied) Score ID slot One 0.810 7402172104936 3 2 0.733 7108222014391 2 3 0.723 6305241032934 One 4 0.715 5812042103428 3 5 0.692 4611052130294 2 6 0.619 7402172104936 2 7 0.618 5812042103428 One 8 0.596 4611052130294 One 9 0.593 6605122130294 One 10 0.506 7204121064591 3

Extracted customer time zone ID slot One 7402172104936 3 2 7108222014391 2 3 6305241032934 One 4 4611052130294 2 5 5812042103428 One

First customer group ID Score_1 Score_2 Score_3 7402172104936 0.493 0.619 0.810 7108222014391 0.448 0.733 0.504 6305241032934 0.723 0.394 0.429 4611052130294 0.596 0.692 0.265 5812042103428 0.618 0.502 0.715

2nd customer group ID Score_1 Score_2 Score_3 7204121064591 0.339 0.218 0.506 6605122130294 0.593 0.215 0.443

By implementing the present invention, it is possible to predict the optimized call time and channel for each customer, and by maximizing the success rate of the call with the customer by attempting to call the customer in the call center according to the predicted result, The cost savings can be achieved.

Claims (38)

M-1) The call recommendation system server receives a call list from the host computer server that stores the call target list, which includes the call recipients and call related information about the call subjects to be performed based on a predetermined unit. ; M-2) the call recommendation system server extracts data about a call target customer on the unit call list from a customer data DB; M-3) the call recommendation system server determines a call success prediction model; M-4) the call recommendation system server assigning a score related to a call for each customer to the determined call success prediction model; M-5) the call recommendation system server to process the call-related score for each customer to specify at least one or more of the call time zone and channel for each customer; M-6) the call recommendation system server, the call recommendation method in the call center system comprising the step of transmitting at least one or more of the specified information of the call time and channel for each customer to the call originating computer server. The method of claim 1, wherein the customer data DB of step M-2) comprises information on call center call history, information on telemarketing (TM) history, customer basic information, credit information, and automated response system (ARS) history. Call recommendation method in a call center system comprising at least one of information, demographic information. The method of claim 1, wherein the method of determining the call success prediction model of the M-3 step is based on case-based reasoning, decision tree, neural network, and spline. Call recommendation method in a call center system, characterized in that using at least one or more of a method including a logistic regression (Logistic Regression) and cluster analysis (Cluster Analysis). The method of claim 1, wherein the method for assigning a score related to a call for each customer in the step M-4) uses a result value obtained by inputting customer information on a customer data DB to a call success prediction model corresponding function. Call recommendation method in call center system. [5] The method of claim 4, wherein the method for assigning a score related to a call for each customer includes at least one method of assigning scores according to time slots, assigning scores for communication media, and assigning scores for receiving locations. How to recommend a call. The method of claim 1, wherein the method for designating a call time zone for each customer by processing the call-related scores for each customer of step M-5) includes at least one of a time zone score, a communication medium score, and a reception place score of the call target customer. A call recommendation method in a call center system, characterized by finding an optimal call connection customer list for each time zone using the above score as an input variable and limiting the maximum number of call customers for each time zone. The method of claim 6, wherein the method of finding the optimal call connection customer list for each time zone uses at least one or more of a grid search method, a linear programming method, and a distance analysis method. The method of claim 6, wherein the call time point for each customer is designated by processing the call-related score for each customer in the step M-5). Extracting a customer list corresponding to the maximum number of call connection customers per time zone as a data structure including a call-related score for each customer; And And sorting the extracted customer list based on at least one or more scores of time points, scores of communication media, and scores of reception locations. The method of claim 1, wherein the step M-4) The call recommendation method of the call center system, characterized in that it further comprises the step of assigning a score including weight information previously given for each customer. 2. The method of claim 1, further comprising: M-7) storing, by the calling computer, the results of the calling through the calling computer; M-8) The call recommendation method of the call center system, characterized in that further comprising the step of reflecting the result of the call originating in step M-7) to the customer data DB. (a) having a database storing predetermined score information previously assigned to each of the M time zones when the IDs of the plurality of customers and the day are divided into a predetermined number of M time zones; (b) if the number of customers is less than or equal to the minimum of the maximum number of customers allowed to connect in each time zone by comparing the minimum value among the maximum number of customers allowed to connect in each time zone, And designating a time zone corresponding to a maximum score among the scores of each customer as the optimum time zone of the customer. The method of claim 11, (c) If the number of customers is greater than the minimum value of the maximum number of customers allowed to make a call in each time period, the sum of the number of customers and the maximum number of customers allowed to make a call in each time period is determined as the result of the step (b). Comparing the number of customers to determine whether or not the number of customers is less than or equal to the sum of the maximum number of customers allowed to be connected to each time zone; (d) If the determination result of step (c) is true, the maximum score is used to match the CNT value, which is the number of customers assigned to each time zone, in one-to-one correspondence with the Max values, the maximum allowable number of customers for each time zone, in that order. Determining whether all CNT values are less than or equal to a corresponding Max value; And (e) if the determination result of step (d) is true, specifying the time zone corresponding to the maximum score among the scores of each customer as the optimum time zone of the customer. How to specify. The method of claim 12, (f) if the result of the determination of step (d) is false, determining whether the M value is greater than two; And (g) if the determination result of step (f) is true, The M time zones correspond to predetermined M axis coordinates, the distance values of the M scores of the respective customers are calculated from the axis coordinates of each time zone, and the optimum time zones are given in order from the smallest distance values. And adjusting the optimal time zone given to each customer based on a difference value between the scores of the two optimal time zones assigned to the optimal time and a rank value which is a rank value for each of the two optimal time zones. How to specify the optimum time zone for each customer, characterized in that. The method of claim 13, wherein step (g) (g1) mapping the M time zones into M axial coordinates as in the following equation; [Equation] A_1 = (1, 0, 0, ..., 0), A_2 = (0, 1, 0, ..., 0) ... A_M = (0, 0, ..., 1), where A_1 , A_2, ..., A_M are coordinate axes of each M time zone. (g2) calculating a distance from a coordinate axis of time zone j for each of the M scores of the customers using the following equation; [Equation 1] Dist _j = {∑ ^M _{k = 1} (Score_k-Kth element of A_j) ² } ^1/2 Where j is any number from 1 to M, M is the number of time zones or scores per customer, and A_j is the axis coordinate of the j time zone. (g3) arranging the obtained distances in ascending order in association with the corresponding customer ID and time zone information; (g4) generating first optimal time information by assigning optimal time zones in order from the smallest distance, except that the optimal time zone is already given or the maximum allowable number of customers for each time zone is already filled; (g5) When the value of the arbitrary variable K is changed from 1 to M and the other arbitrary variable L is changed from 1 to M, the optimum time zone is excluded, except that the value of K and the value of L are the same. Extracting, from the first optimal time information, an ID of an adjustment target customer who is K or a customer such as L, a K th score and an L th score of the customer having the ID; (g6) calculating a difference value and a Rank (R) between two scores of the adjustment target customer; And (g7) When the number of the adjustment target customers is N, the difference between the scores of the j th customer is D_j, and the rank (R) of the j th customer is R_j, the optimal time zone of the j th customer is the R_j. Is less than or equal to Max_K and D_j is greater than or equal to 0, the optimal time zone for each customer, characterized by being given as the K-th time zone. The method of claim 14, wherein step (g7) If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is less than or equal to Max_L, the optimal time zone for each customer is assigned. The method of claim 14, wherein step (g7) If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is greater than Max_L, the optimal time zone for each customer is assigned. The method of claim 14, (g8) further comprising repeating steps (g5) to (g7) until the optimal time zone of all the customers no longer changes. The method of claim 13, wherein step (g) (g11) calculating a difference value and a rank value of scores for the two time zones of each customer when the determination result of step (f) is false; And (g12) N is the total number of customers, and the difference value of the score calculated in step (g1) of the variable j-th customer having an arbitrary value is D_j, and the calculated rank value of the j-th customer is R_j. In this case, the optimal time zone of the j th customer includes the step of designating the first time zone when the R_j value is less than or equal to Max_1 and the D_j value is greater than or equal to 0. Way. 19. The method of claim 18, wherein step (g12) When the total number of customers is N, and the difference value of the score calculated in the step (g11) of the variable j-th customer having an arbitrary value is D_j and the calculated rank value of the j-th customer is R_j, The optimum time zone of the j-th customer is specified as the second time zone when the difference value obtained by subtracting the R_j value from the N value is less than or equal to Max_2. 19. The method of claim 18, wherein step (g12) When the total number of customers is N, and the difference value of the score calculated in the step (g1) of the variable j-th customer having an arbitrary value is D_j and the calculated rank value of the j-th customer is R_j, The optimum time zone of the j-th customer is specified as the first time zone when the difference value obtained by subtracting the R_j value from the N value is greater than Max_2. The method of claim 12, In step (d), if the result of the determination in step (c) is false, the scores of the customers are lowered in descending order until the number of allowed customers in each time slot is filled, except for customers assigned to any time slot. While the customer ID and time zone information is sorted, the count (CNT) of the time zone is increased by 1, and the first customer set, which is a set of customers assigned to each of the time zones, and a second customer set not assigned to any time zone, are included. Obtaining; And In the step (e), when all the time (CNT) values of each time zone and the time values of the Max (CNT) correspond one-to-one, it is determined whether all the count (CNT) values are smaller than or equal to the corresponding Max value, and the first value is true. And specifying a time zone corresponding to a maximum score among the scores of each customer belonging to a customer set as the optimum time zone of the customer. The method of claim 12, In step (e), when the CNT values of each time zone and the Max values of the time zone correspond one-to-one, whether all CNT values are smaller than or equal to the corresponding Max value is false. Judge whether or not, (f) calculating a difference value and a rank value of scores for two time zones of each customer belonging to the first set of customers if the determination result of step (e) is false; (g) In step (d), the total number of customers belonging to the first customer set is calculated in step (f) of the variable j-th customer having any value of N and customers belonging to the first customer set. When the difference value of the score is D_j and the calculated rank value of the j th customer is R_j, the optimal time zone of the j th customer is equal to or smaller than Max_1 and the D_j value is greater than or equal to 0. If the first time zone; (h) When the value of the random variable K varies from 1 to M, the customers belonging to the first customer set are extracted the customers with the Kth time zone assigned to the optimal time zone and the customers of the second customer set not assigned the optimal time zone. When the extracted customers are sorted in descending order based on the scores of the extracted K customers' time zones, and the score Score_K rank of the Kth time zones of the j th customers, which is a random value among the sorted customers, is R_j. If the R_j value is less than or equal to the Max_K value, the optimal time zone of the first customer designates the K-th time zone as the optimal time; otherwise, the optimal time zone of the j-th customer is not specified; (i) When the value of any variable L varies from 1 to M, except for the case where the L value is equal to the K value, among the customers belonging to the first customer set, the L time zone is assigned to the optimal time zone, After extracting the customers of the second unassigned customer set, the extracted customers are sorted in descending order based on the scores of the extracted L customers' time zones, and the K th customer of the j th customer is any value of the sorted customers. When the score Score_L rank of the time zone is R_j, the optimal time zone of the j th customer is designated as the L time zone as the optimal time when the R_j value is less than or equal to the Max_L value. Not specifying; (j) extracting, from the first optimal time information, an ID of the adjustment target customer whose optimal time zone is K or a customer such as L, the Kth score and the Lth score of the customer having the ID, Calculating a difference value and Rank (R) between two scores; And (k) If the number of customers to be adjusted is N, the difference between the scores of the j th customer is D_j, and the rank (R) of the j th customer is R_j, the optimal time zone of the j th customer is R_j. Is less than or equal to Max_K and D_j is greater than or equal to 0, the method further comprising the step of assigning the K-th time zone. The method of claim 22, wherein step (k) If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is less than or equal to Max_L, the optimal time zone for each customer is assigned. The method of claim 22, wherein step (k) If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is greater than Max_L, the optimal time zone for each customer is assigned. The method of claim 22, wherein step (g) When the total number of customers is N, and the difference value of the score calculated in the step (f) of the variable j-th customer having an arbitrary value is D_j, and the calculated rank value of the j-th customer is R_j, The optimum time zone of the j-th customer is specified as the second time zone when the difference value obtained by subtracting the R_j value from the N value is less than or equal to Max_2. The method of claim 22, wherein step (g) When the total number of customers is N, and the difference value of the score calculated in the step (f) of the variable j-th customer having an arbitrary value is D_j, and the calculated rank value of the j-th customer is R_j, The optimum time zone of the j-th customer is specified as the first time zone when the difference value obtained by subtracting the R_j value from the N value is greater than Max_2. The method of claim 22, and (l) repeating steps (h) to (k) until the optimal time zone of all the customers no longer changes. The method of claim 21, Step (e) determines whether all the CNT values are less than or equal to the Max value corresponding to each time when the CNT values for each time zone and the Max values for each time zone correspond one-to-one, (f) determining whether the value of M is greater than 2 when the determination result of step (e) is true; (g) if the determination result of step (f) is true, The M time zones correspond to predetermined M axis coordinates, and the distance values from the M coordinates of the respective time zones are calculated for M scores of the customers belonging to the first customer set. And a value assigned to each customer belonging to the first customer set based on a difference value and a Rank (R) value between the scores of the two optimum time bands of each customer belonging to the first customer set given to any two optimal times. Adjusting an optimum time zone; (h) When the value of the random variable K varies from 1 to M, the customers belonging to the first customer set are extracted the customers with the Kth time zone assigned to the optimal time zone and the customers of the second customer set not assigned the optimal time zone. When the extracted customers are sorted in descending order based on the scores of the extracted K customers' time zones, and the score Score_K rank of the Kth time zones of the j th customers, which is a random value among the sorted customers, is R_j. If the R_j value is less than or equal to the Max_K value, the optimal time zone of the first customer designates the K-th time zone as the optimal time; otherwise, the optimal time zone of the j-th customer is not specified; (i) When the value of any variable L varies from 1 to M, except for the case where the L value is equal to the K value, among the customers belonging to the first customer set, the L time zone is assigned to the optimal time zone, After extracting the customers of the second unassigned customer set, the extracted customers are sorted in descending order based on the scores of the extracted L customers' time zones, and the K th customer of the j th customer is any value of the sorted customers. When the score Score_L rank of the time zone is R_j, the optimal time zone of the j th customer is designated as the L time zone as the optimal time when the R_j value is less than or equal to the Max_L value. Not specifying; (j) extracting, from the first optimal time information, an ID of the adjustment target customer whose optimal time zone is K or a customer such as L, the Kth score and the Lth score of the customer having the ID, Calculating a difference value and Rank (R) between two scores; And (k) If the number of customers to be adjusted is N, the difference between the scores of the j th customer is D_j, and the rank (R) of the j th customer is R_j, the optimal time zone of the j th customer is R_j. Is less than or equal to Max_K and D_j is greater than or equal to 0, the method further comprising the step of assigning the K-th time zone. The method of claim 28, wherein step (k) If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is less than or equal to Max_L, the optimal time zone for each customer is assigned. The method of claim 28, wherein step (k) If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is greater than Max_L, the optimal time zone for each customer is assigned. The method of claim 28, and (l) repeating steps (h) to (k) until the optimal time zone of all the customers no longer changes. The method of claim 28, wherein step (g) (g1) mapping the M time zones into M axial coordinates as in the following equation; A_1 = (1, 0, 0, ..., 0), A_2 = (0, 1, 0, ..., 0) ... A_M = (0, 0, ..., 1), where A_1 , A_2, ..., A_M are coordinate axes of each M time zone. (g2) calculating a distance from a coordinate axis of time zone j for each of the M scores of the customers using the following equation; [Equation 1] Dist _{j =} {∑ ^M _{k = 1} (Score_k-Kth element of A_j) ² } ^1/2 Where j is any number from 1 to M, M is the number of time zones or scores per customer, and A_j is the axis coordinate of the j time zone. (g3) arranging the obtained distances in ascending order in association with the corresponding customer ID and time zone information; (g4) generating first optimal time information by assigning optimal time zones in order from the smallest distance, except that the optimal time zone is already given or the maximum allowable number of customers for each time zone is already filled; (g5) When the value of the arbitrary variable K is changed from 1 to M and the other arbitrary variable L is changed from 1 to M, the optimum time zone is excluded, except that the value of K and the value of L are the same. Extracting, from the first optimal time information, an ID of an adjustment target customer who is K or a customer such as L, a K th score and an L th score of the customer having the ID; (g6) calculating a difference value and a Rank (R) between two scores of the adjustment target customer; And (g7) When the number of the adjustment target customers is N, the difference between the scores of the j th customer is D_j, and the rank (R) of the j th customer is R_j, the optimal time zone of the j th customer is the R_j. Is less than or equal to Max_K and D_j is greater than or equal to 0, the optimal time zone for each customer, characterized by being given as the K-th time zone. The method of claim 32, wherein step (g7) If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is less than or equal to Max_L, the optimal time zone for each customer is assigned. The method of claim 32, wherein step (g7) If the number of customers to be adjusted is N, the difference between the scores of the j-th customer is D_j, and the rank (R) of the j-th customer is R_j, the optimal time zone of the j-th customer is N-R_j + If the value of 1 is greater than Max_L, the optimal time zone for each customer is assigned. 33. The method of claim 32, (g8) further comprising repeating steps (g5) to (g7) until the optimal time zone of all the customers no longer changes. 36. The method of claim 1, wherein the call recommendation system server of step M-5 processes the call-related scores for each customer to designate a call time slot for each customer. Call recommendation method in a call center system, characterized in that applied. A recording medium having recorded thereon a program capable of executing the method of any one of claims 1 to 35 with a computer. A call center system equipped with a program capable of executing the method of any one of claims 1 to 35.