JP5094643B2 - Expected successful bid price calculation apparatus, expected successful bid price calculation method, and computer program - Google Patents

Description

  The present invention relates to an expected successful bid price calculation apparatus, an expected successful bid price calculation method, and a computer program for predicting the highest bid price (successful bid price) in a bidding system in which a plurality of persons bid for goods.

  FIG. 11 is a conceptual diagram showing the concept of a conventional auction mechanism. As shown in FIG. 11, in the conventional auction, a provider (supplier) puts a product for sale, and a plurality of traders (buyers) who want to purchase the product bid each, and the highest bid is obtained. A buyer who presents a price (successful bid price) purchases the product. Accordingly, each buyer cannot know the number of bids and the bid amount by other buyers at the time of bidding. However, for each buyer, if it is possible to know the number of bids and the bid amount by other buyers, it is possible to consider the bid based on the information and make a bid advantageously. Therefore, the number of bids and bid amounts by other buyers are valuable information for each buyer.

  In response to such needs of each buyer, a service has been proposed in which a successful bid price is predicted based on a past successful bid history and the expected successful bid price is provided to the buyer. Specifically, a model that predicts the winning bid price using information (for example, model number, performance, good / bad condition, etc.) related to the product for sale is constructed based on the past winning bid history, and the expected winning bid price. A technique for calculating the value has been proposed.

For example, the technology disclosed in Patent Document 1 can calculate the probability distribution of successful bid prices of properties in an auction in an auction in which real estate is handled as a product, and can simulate the transition of the highest bid price.
JP 2008-070917 A

  However, in the technique disclosed in Patent Document 1 described above, a virtual buyer on a computer called an agent repeatedly decides a bid amount based on the highest bid price at that time and repeatedly performs a bid. The predicted value of the final highest bid price (successful bid price) is calculated. Therefore, there is a problem that it is difficult to calculate a predicted value for an auction based on simultaneous bidding.

  In view of the above circumstances, an object of the present invention is to provide an expected successful bid price calculation apparatus that can predict a successful bid price in an auction based on simultaneous bidding and calculate an expected successful bid price.

[1] In order to solve the above-described problem, an expected successful bid price calculation apparatus according to one aspect of the present invention provides, for each bid item, product information, product type, successful bid price, and number of bids for past bid target products. And a successful bid history database that preliminarily stores successful bid history data that associates successful buyers, and for each of the same type of products, based on the successful bid history data, the number of bids and product information are explanatory variables, and the winning bid is a target variable The prediction formula calculation unit for calculating the parameters of the prediction formula of the winning bid price by performing the multiple regression analysis, the clustering unit for clustering the winning bid history data for each item, and the product information of the product to be newly bidd Is calculated as a probability of being included in each class of the clustered buyer and an average of the number of successful bids of the product type in each class of buyer Calculated and, by adding the multiplication results in the class of all buyers, the expected bid number calculating unit for calculating a number of expected bid for products made with the new bid target, the prediction expression, and the estimated bids, An expected successful bid price calculation unit that calculates an expected successful bid price of the product to be newly bidd by applying the product information of the product to be newly bidd;

  Further, one aspect of the present invention may be specified as an expected successful bid price calculation method performed by the above-described expected successful bid price calculation apparatus. Further, one aspect of the present invention may be specified as a computer program for causing a computer including the above-described successful bid history database to execute the above method.

  According to the present invention, it is possible to calculate an expected successful bid price in an auction based on simultaneous bidding.

  FIG. 1 is a conceptual diagram showing a concept of an auction mechanism using a bidding support information providing apparatus (corresponding to “predicted bidding price calculation apparatus” in the present invention). In this auction mechanism, a supplier provides a product, and information related to the product is input by a bid support information providing apparatus. Next, the bid support information providing device predicts the number of bids and the winning bid price, and provides the expected number of bids and the expected winning bid price (hereinafter, these two are collectively referred to as “bid supporting information”) to the buyer. Then, the buyer bids based on the bid support information.

  Hereinafter, the bidding support information providing apparatus 1 when the commodity to be traded is a used car will be described. FIG. 2 is a functional block diagram showing a functional configuration of a bid support information providing apparatus 1 that predicts bid support information and provides it to buyers in a used car auction. As shown in the figure, the bid support information providing apparatus 1 includes a successful bid history DB 101, a vehicle type history DB 102, a vehicle type history creation unit 103, a prediction formula calculation unit 104, a buyer history DB 105, a buyer history creation unit 106, a clustering DB 107, a clustering unit 108, An expected bid number calculation unit 109 and an expected successful bid price calculation unit 110 are provided.

The successful bid history DB 101 previously stores past successful bid history data in a used car auction.
The vehicle type history DB 102 stores vehicle type history data extracted for each vehicle type from the successful bid history data stored in the successful bid history DB 101.

The vehicle type history creation unit 103 extracts vehicle type history data from successful bid history data stored in the successful bid history DB 101 and writes the extracted vehicle type history data to the vehicle type history DB 102.
The prediction formula calculation unit 104 calculates the value of each parameter of the prediction formula used in the predicted successful bid price calculation unit 110 based on the vehicle type history data stored in the vehicle type history DB 102.

The buyer history DB 105 stores buyer history data extracted for each buyer from the successful bid history data stored in the successful bid history DB 101.
The buyer history creation unit 106 extracts the buyer history data from the successful bid history data stored in the successful bid history DB 101 and writes it to the buyer history DB 105.

The clustering DB 107 stores clustering data and class correspondence data, which are the results of clustering buyer, vehicle type, and multiple vehicle information.
Based on the buyer history data, the clustering unit 108 creates clustering data and class correspondence data based on each of the buyer, the vehicle type, and the plurality of vehicle information, and writes the clustering data and class correspondence data in the clustering DB 107.

The predicted number-of-bids calculation unit 109 applies to the used vehicles to be bid based on the clustering data stored in the clustering DB 107, the class correspondence data, and the vehicle information (bidding target vehicle information) of the used vehicles to be bid. Calculate the expected number of bids (expected number of bids).
The predicted successful bid price calculation unit 110 is based on the predicted bid number calculated by the predicted bid number calculation unit 109 and the value of each parameter of the prediction formula calculated by the prediction formula calculation unit 104. The expected price of the successful bid price (expected successful bid price) is calculated.

  FIG. 3 is a schematic diagram showing an outline of successful bid history data stored in advance by the successful bid history DB 101. The successful bid history data includes, for each bidding item, the buyer ID of the buyer who has made a successful bid, the model number indicating the model of the used car that has been awarded, the winning bid price, the number of bids, and multiple pieces of vehicle information (vehicle information 1 to The vehicle information N) is associated with the data. The vehicle information is information relating to, for example, the total length of the vehicle, the vehicle width, the vehicle height, the vehicle body weight, the travel distance, the number of years, and the presence or absence of scratches. Also, binary information such as the presence or absence of a flaw is expressed using a binary representation (binary representation) of “0” or “1”. For example, “1” indicates that there is a flaw and “0” indicates that there is no flaw.

  FIG. 4 is a schematic diagram showing an outline of vehicle type history data stored by the vehicle type history DB 102. The vehicle type history data is data in which the winning bid price, the number of bids, and a plurality of pieces of vehicle information (vehicle information 1 to N) are associated with each vehicle type for each vehicle type. The vehicle type history data is created by the vehicle type history creating unit 103 and written in the vehicle type history DB 102.

  FIG. 5 is a schematic diagram showing an outline of buyer history data stored by the buyer history DB 105. The buyer history data is data in which for each buyer ID, the number of successful bids for each vehicle type (vehicle type 1 to vehicle type L), the average successful bid price, and the average value of each vehicle information related to used vehicles that have been successfully bid in the past are associated. The buyer history data is created by the buyer history creating unit 106 and written in the buyer history DB 105. In the buyer history data, the value of the vehicle information represented using binary representation may be a value rounded to “0” or “1”, or may be a decimal value.

  FIG. 6 is a schematic diagram showing an outline of clustering data stored by the clustering DB 107. The clustering data is data in which for each buyer class number, the number of successful bids for each vehicle type class, the average value of successful bid prices through all vehicle types, and the average value of each vehicle information class are associated. The number of successful bids for each vehicle type class represents the sum of the number of successful bids for each vehicle type included in that vehicle type class among the number of successful bids made by each buyer included in the corresponding buyer class. The average value of the winning bids represents the average value of all winning bids made by the buyers included in the corresponding buyer class regardless of the vehicle type class. The average value of each vehicle information class represents the average value for each vehicle information included in the vehicle information class, out of the number of vehicle information related to vehicles that have been awarded by buyers included in the corresponding buyer class. Clustering data is created by the clustering unit 108 and written in the clustering DB 107.

  FIG. 7 is a schematic diagram showing an outline of the class correspondence data stored by the clustering DB 107. The class correspondence data includes buyer class correspondence data (FIG. 7A), vehicle type class correspondence data (FIG. 7B), and vehicle information class correspondence data (FIG. 7C). The buyer class correspondence data is data in which a buyer class number is associated with each buyer ID. The vehicle class correspondence data is data in which a vehicle type class number is associated with each vehicle type number. The vehicle information correspondence data is data in which a vehicle information class number is associated with each vehicle information number. The class correspondence data is created by the clustering unit 108 and written into the clustering DB 107.

  FIG. 8 is a flowchart showing an operation example of the bid support information providing apparatus 1. Hereinafter, an operation example of the bid support information providing apparatus 1 will be described with reference to FIG. First, the vehicle type history creation unit 103 reads successful bid history data from the successful bid history DB 101, creates vehicle type history data, and writes it in the vehicle type history DB 102 (step S101).

  Next, the prediction formula calculation unit 104 reads the vehicle type history data from the vehicle type history DB 102, and calculates the value of each parameter of the prediction formula for every vehicle type for all vehicle types (step S102). Specifically, the prediction formula calculation unit 104 performs a multiple regression analysis with the bid number and each of the vehicle information 1 to N as explanatory variables for a certain vehicle type and the winning bid as a target variable. The value of each parameter used in the prediction formula for calculating the predicted successful bid price is calculated. The prediction formula is, for example, Formula 1, and each parameter is, for example, a_1 to a_ {N + 1} and b in Formula 1. At this time, the prediction formula calculation unit 104 may select a variable based on the stepwise method.

  FIG. 9 is a schematic diagram illustrating an outline of each parameter calculated by the prediction formula calculation unit 104. As shown in FIG. 9, the prediction formula calculation unit 104 performs the above-described processing on all vehicle types, thereby increasing the number of bids for every vehicle type (that is, for each vehicle type number). And the value (including the intercept) of each parameter related to each vehicle information is calculated.

  Next, the buyer history creation unit 106 reads the successful bid history data from the successful bid history DB 101, creates buyer history data, and writes it to the buyer history DB 105 (step S103).

  Next, the clustering unit 108 reads the buyer history data from the buyer history DB 105, performs clustering to create clustering data and class correspondence data, and writes the clustering data and class correspondence data to the clustering DB 107 (step S104). Details of processing performed by the clustering unit 108 will be described later.

  When the processing of the clustering unit 108 ends, the predicted bid number calculation unit 109 reads the clustering data from the clustering DB 107 and calculates the predicted bid number for the vehicle to be bid (step S105). Note that the details of the process performed by the predicted bid number calculation unit 109 will be described later, as in the process performed by the clustering unit 108.

  When the process of the predicted bid number calculation unit 109 is completed, the predicted successful bid price calculation unit 110 calculates the prediction formula calculated by the prediction formula calculation unit 104, the predicted bid number calculated by the predicted bid number calculation unit 109, An expected successful bid price is calculated based on the model of the target vehicle and the vehicle information 1 to N (step S106).

  FIG. 10 is a schematic diagram showing an outline of information (bidding target vehicle information) input to the bidding support information providing apparatus 1 as information on a vehicle to be bid. The input target vehicle information includes the model number of the vehicle to be bid and vehicle information 1 to N. Specifically, the predicted successful bid price calculation unit 110 includes information on the vehicle to be bid, parameters of the prediction formula (parameters shown in FIG. 9) corresponding to the model of the vehicle to be bid, Is used to calculate the expected successful bid price. Then, the predicted successful bid price calculation unit 110 outputs the predicted number of bids and the predicted successful bid price (that is, bid support information) (step S107), and ends the processing of the entire flowchart.

Next, details of processing performed by the clustering unit 108 will be described. The successful bid history data is regarded as a vector of D = {D P , D T , D R , D B , D C }. D P is the buyer i (i = 1, 2, · · ·, N) with the average bid price d i (P) is the i element of a vector of N dimensions. D T has vehicle type t buyer i is awarded in the past (t = 1,2, ···, L T) number d i of, t a (T) to the (i, t) components, N × L T matrix. D R is the car information (continuous values) relating to car buyers i is awarded in the past j (j = 1,2, ···, L R) Mean value d i, j a (R) a (i in, j ) The element has an N × LR matrix. Here, the vehicle information (continuous value) is a continuous value of the vehicle information, and is vehicle information expressed by a continuous value such as the total length of the vehicle or the weight of the vehicle body. D B is vehicle information about vehicles buyer i is awarded in the past (binary value) k (k = 1,2, ··· , L B) the average value d i of, k and (B) the (i, k ) with the element, an N × L B matrix. Here, the vehicle information (binary value) is a binary value of the vehicle information. For example, the vehicle information represented by binary values (0, 1) such as the presence or absence of vehicle inspection or the presence or absence of scratches. It is. The average value d i, k (B) is a value rounded to 0 or 1 by rounding off. D C is vehicle information (count value) relating to car buyers i is awarded in the past l (l = 1,2, ···, L C) the average value d i of, l and (C) the (i, l ) with the element, an N × L C matrix. Here, the vehicle information (count value) is the vehicle information whose value is a count value, and is vehicle information expressed by a frequency (positive number) such as the number of registered months and the total number of scratches, for example. . The average value d i, l (C) is a value rounded to an integer value by rounding off.

The clustering unit 108 clusters the buyer, vehicle type, vehicle information (continuous value), vehicle information (binary value), and vehicle information (count value). Here, a vector “class assignment vector A” in which class numbers are arranged for each clustering target is defined as A = {Z, W _T , W _R , W _B , W _C }. Z is an N-dimensional vector having buyer class z _i to which buyer i belongs as the i-th element. W _T is a _{L T-dimensional} vector with the model class number _{w t} belonging models t to t-th element. W _R is the affiliation to vehicle information (continuous values) class number w _j Features j is L _{R-dimensional} vector with the j-th element. W _B is the affiliation to vehicle information (binary value) class number w _k characteristic k is L _B-dimension vector with the k-th element. W _C is belonging to vehicle information (count value) class number w _l Features l is L _{C-dimensional} vector with the first l elements. In this way, the number of different vector elements is the number of classes. For example, if the number of buyers N = 5, the number of classes is 4 when Z = [1, 2, 4, 3, 1], and the class when Z = [2, 2, 2, 1, 1]. The number is 2.

  First, the clustering unit 108 randomly determines the number of classes for each buyer, each vehicle type, each vehicle information (continuous value), each vehicle information (binary value), and each vehicle information (count value), A class number is randomly assigned to each data. Next, the clustering unit 108 selects one data (for example, one buyer, one car type, one car information, etc.) and assigns a new class number to the selected data. Update the class number assigned to the data. A method for selecting a new class number will be described later. Then, after the update, the clustering unit 108 calculates the posterior probability of the class assignment vector A. The clustering unit 108 repeatedly executes these processes for other data, and compares the values of the posterior probabilities calculated each time. Then, when a predetermined condition, for example, the value of the posterior probability has not risen above the threshold continuously for a predetermined number of times, or when the value of the posterior probability exceeds the threshold, the clustering unit 108 ends this iterative process, The clustering is determined by the class number assignment at the time of completion.

Here, the infinite relationship model outputs A ^* that maximizes the appearance probability of the class assignment vector A under the given bidding history, that is, the posterior probability of the class assignment vector A. At this time, A ^* is expressed as shown in Equation 2.

  Further, the posterior probability of the class assignment vector can be expanded as shown in Equation 3 under the assumed model.

  The assumed distribution is set in advance for each data included in the successful bid history. Specifically, distributions as in Expressions 4 to 8 are set in advance.

Here, Θ _P , Θ _T , Θ _R , Θ _B , and Θ _C represent parameters for the respective distributions. Specifically, Equation 9 represents normal distribution parameters (mean and variance) in the buyer class r.

  Expression 10 represents the parameters of the multinomial distribution in the buyer class r and the vehicle type class s.

  Equation 11 represents the parameters (average and variance) of the normal distribution in the buyer class r and the vehicle information (continuous value) class s.

  Expression 12 represents a Bernoulli distribution parameter in the buyer class r and the vehicle information (binary value) class s.

  Expression 13 represents a Poisson distribution parameter in the buyer class r and the vehicle information (count value) class s.

Note that the prior distributions p (Θ _P ), p (Θ _T ), p (Θ _R ), p (Θ _B ), and p (Θ _C ) are set as the distribution parameters, respectively. Similarly, a prior distribution is set for the class assignment vector A. For example, it is possible to set each Dirichlet Process prior distribution.

  Here, α, βT, βR, βB, and βC represent the parameters of the Dirichlet Process prior distribution.

The Gibbs sampler method can be used as a method of calculating class number vectors Z ^* , W _T ^* , W _R ^* , W _B ^* , and W _C ^* that maximize the posterior probability. In this case, the posterior probability can be maximized by updating the class number vector one by one. For example, when updating the buyer class number zi of the buyer i, first, the probability that the buyer i belongs to each buyer class c (c = 1, 2,...) Is calculated using Equation 19.

  The clustering unit 108 calculates the probability using Expression 19 when updating the assignment of the class number to certain data. Next, the clustering unit 108 sets the probability that each buyer class number appears as the probability calculated by Expression 19, and updates the newly assigned buyer class number by assigning it to the buyer i. For example, the clustering unit 108 sets a virtual dice in which the probability of each “eye” appearing matches the probability calculated by Equation 19, and sets the “eye” (buyer class number) generated by rolling the dice to the buyer. Newly assigned to i.

In the same procedure, the vehicle type class number w _t , the vehicle information (continuous value) class number w _j , the vehicle information (binary value) class number w _k , and the vehicle information (count value) class number w _t are updated. For example, Expression 20 is used for updating the vehicle class number w _t , and Expression 21 is used for updating the vehicle information (continuous value) class number w _j .

The order of updating the class numbers may be appropriately set by the designer. For example, the buyer class number zi, the vehicle type class number w _t , the vehicle information (continuous value) class number w _j , the vehicle information (binary value) class number w _k and vehicle information (count value) may be set to be updated one by one in the order of the class number w _t .

Next, details of the processing of the expected bid number calculation unit 109 will be described. Let t be the vehicle type of the vehicle that is the subject of the bid (hereinafter referred to as “exhibited vehicle”), and the vehicle information (continuous value), vehicle information (binary value), and vehicle information (count value) are x ^(R) , Let x ^(B) and x ^(C) . Then, the similarity between the vehicle that the buyer class r has made a successful bid in the past and the exhibited vehicle can be calculated as the probability that the attribute information of the exhibited vehicle appears in the buyer class r as shown in Equation 22.

  The larger the value calculated in Equation 22, the more similar the vehicle that the buyer class r has made a successful bid in the past and the exhibited vehicle. Therefore, it can be said that the larger the calculated value is, the higher the possibility of bidding by the buyer class r is. Therefore, the expected number of bids can be expressed as in Expression 23.

Here, U represents the number of buyer classes, and the value 24 represents the average number of successful bids for the vehicle type t in the buyer class r. The values 25 to 27 can be calculated from the probability distribution for each data set in advance and the model parameters Θ ^(R) , Θ ^(B) , and Θ ^(C) obtained as a result of clustering. . For example, the value 25 can be calculated by Equation 28.

  The predicted bid number calculation unit 109 calculates the predicted bid number by using the number 22 in this way.

  The bid support information providing apparatus 1 configured in this way calculates the similarity between the vehicle information and the exhibited vehicle that each buyer class has made a successful bid for, calculates the expected number of bids in each buyer class according to this similarity, The expected number of bids is calculated by adding the expected number of bids in the buyer class. Based on the predicted number of bids, an expected successful bid price is calculated. Therefore, it is not necessary to repeat simulation of bidding by a virtual buyer as in the past, and it is possible to calculate an expected successful bid price even in an auction based on simultaneous bidding.

  In addition, the expected bid number calculation unit 109 in the bid support information providing apparatus 1 configured as described above causes the attribute information of the exhibited vehicle in each buyer class to appear in the process of calculating the expected bid number by the process using Expression 22. The probability of doing is calculated. Therefore, the expected successful bid price calculation unit 110 of the bid support information providing apparatus 1 may be configured to provide bid support information only to buyers included in the buyer class having a high value. By being configured in this way, it becomes possible to provide bid support information only to buyers who are likely to actually need this bid support information, and provide services according to buyer's preference. , Buyers' satisfaction can be improved.

  Further, the bid support information may be configured to further provide information on other buyers included in the buyer class having a high value (value calculated by the above-described equation 22). This configuration allows us to provide information on whether other buyers are likely to bid, and which buyers are more likely to bid, providing more detailed bidding assistance. It becomes possible.

  The explanation about the bid support information providing device 1 described above is a description of the configuration when the traded product is a used car, but the bid support information providing device 1 provides bid support information about other products. It may be configured to. In that case, the successful bid history DB 101 is configured to store in advance past successful bid history data for a product to be traded. Also, instead of the vehicle information, specification information about the product to be traded and information about the quality and state are used. Then, the bid support information providing device 1 outputs the expected number of bids and the expected successful bid price for the product to be bid.

  Further, in the bid support information providing apparatus 1 described above, clustering is performed by providing the clustering unit 108, and the expected number of bids is calculated based on the clustered data, but the clustering unit 108 is not provided and clustering is performed. The number of expected bids may be calculated on the basis of missing data (buyer history data).

You may make it implement | achieve the function of the bid assistance information provision apparatus 1 in embodiment mentioned above with a computer. In that case, the program for realizing the function of the bidding support information providing apparatus 1 is recorded on a computer-readable recording medium, and the program recorded on the recording medium is read into the computer system and executed. Also good. Here, the “computer system” includes an OS and hardware such as peripheral devices. The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory inside a computer system serving as a server or a client in that case may be included and a program that holds a program for a certain period of time. The program may be a program for realizing a part of the functions described above, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system.
The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes designs and the like that do not depart from the gist of the present invention.

It is a conceptual diagram showing the concept of the structure of the auction using a bid assistance information provision apparatus. It is a functional block diagram showing the functional structure of the bid assistance information provision apparatus which provides bid assistance information in a used car auction. It is the schematic showing the outline of successful bid history data memorized beforehand by successful bid history DB. It is the schematic showing the outline of the vehicle type log | history data memorize | stored by vehicle type log | history DB. It is the schematic showing the outline of the buyer history data memorize | stored by buyer history DB. It is the schematic showing the outline of the clustering data memorize | stored by clustering DB. It is the schematic showing the outline of the class corresponding | compatible data memorize | stored by clustering DB. It is a flowchart showing the operation example of a bid assistance information provision apparatus. It is the schematic showing the outline of each parameter computed by a prediction formula calculation part. It is the schematic showing the outline of the information input into a bid assistance information provision apparatus as information regarding the vehicle used as a bid object. It is a conceptual diagram showing the concept of the structure of the conventional auction.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Bid assistance information provision apparatus, 101 ... Successful bid history DB, 102 ... Vehicle type history DB, 103 ... Vehicle type history creation part, 104 ... Prediction formula calculation part, 105 ... Buyer history DB, 106 ... Buyer history creation part, 107 ... Clustering DB, 108 ... clustering unit, 109 ... expected bid number calculation unit, 110 ... expected bid price calculation unit

Claims (3)

Every bidding projects, and product information in the past of a bid target products, and the type of goods, and the winning bid, and the bid number, and a successful bid history database for storing in advance the successful bid history data that associates and buyers by which it was awarded,
Prediction formula for calculating parameters of prediction formula of winning bid price by performing multiple regression analysis with the number of bids and product information as explanatory variables and winning bid price as objective variable based on the winning bid history data for the same kind of products A calculation unit;
A clustering unit that clusters the successful bid history data for each item;
Multiplying the similarity calculated as the probability that the product information of the product to be newly tendered is included in each class of the clustered buyer by the average number of successful bids of the product type in each class of buyer, An expected bid number calculation unit for calculating an expected bid number for the new product to be bid by adding the multiplication results in all buyer classes ;
An expected successful bid price calculation unit that calculates an expected successful bid price of the product to be newly bid by applying the predicted number of bids and product information of the new product to be bid to the prediction formula;
An expected successful bid price calculation apparatus. For each bid projects, expected a successful bid with the product information of the past of the bid subject goods, and the type of goods, and the winning bid, and the bid number, the bid history database for storing in advance the successful bid history data that associates and buyers by which it was awarded The price calculation device performs a parameter analysis of the winning bid price by performing multiple regression analysis using the winning bid price and the objective variable as the explanatory variable with the number of bids and the product information as the explanatory variable based on the winning bid history data for each of the same type of products. Calculating steps,
The expected bidding price calculation device clustering the bidding history data for each item;
The predicted successful bid price calculation device calculates the degree of similarity calculated as the probability that the product information of the product to be newly tendered is included in each class of the clustered buyer, and the successful bid of the type of the product in each class of the buyer Multiplying the average of the numbers and adding the multiplication results in all buyer classes to calculate the expected number of bids for the new bid item;
The predicted successful bid price calculation device calculates the predicted successful bid price of the product to be newly bid by applying the predicted number of bids and the product information of the new product to be bid to the prediction formula. And steps to
The method of calculating the expected winning bid price. Every bidding projects, and product information in the past of a bid target products, and the type of goods, and the winning bid, and the bid number, the computer having a successful bid history database for storing in advance the successful bid history data that associates and buyers by which it was awarded In contrast,
Calculating parameters of a prediction formula for a winning bid price by performing a multiple regression analysis with the number of bids and the merchandise information as explanatory variables and the winning bid price as a target variable based on the winning bid history data for each same type of merchandise; ,
Clustering the successful bid history data for each item;
Multiplying the similarity calculated as the probability that the product information of the product to be newly tendered is included in each class of the clustered buyer by the average number of successful bids of the product type in each class of buyer, Calculating the expected number of bids for the newly bidd product by adding the multiplication results in all buyer classes ;
Calculating the expected successful bid price of the new bid target product by applying the predicted number of bids and the product information of the new bid target product to the prediction formula;
A computer program for running.

Images