[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

CN116662564A - Service recommendation method based on depth matrix decomposition and knowledge graph - Google Patents

Service recommendation method based on depth matrix decomposition and knowledge graph Download PDF

Info

Publication number
CN116662564A
CN116662564A CN202310584709.XA CN202310584709A CN116662564A CN 116662564 A CN116662564 A CN 116662564A CN 202310584709 A CN202310584709 A CN 202310584709A CN 116662564 A CN116662564 A CN 116662564A
Authority
CN
China
Prior art keywords
entity
service
vector
user
service item
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202310584709.XA
Other languages
Chinese (zh)
Inventor
付春雷
吴冕
唐鹏辉
李成高
洪伟
赵义伟
鄢萌
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing University
Original Assignee
Chongqing University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Chongqing University filed Critical Chongqing University
Priority to CN202310584709.XA priority Critical patent/CN116662564A/en
Publication of CN116662564A publication Critical patent/CN116662564A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/95—Retrieval from the web
    • G06F16/953—Querying, e.g. by the use of web search engines
    • G06F16/9535—Search customisation based on user profiles and personalisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Business, Economics & Management (AREA)
  • Databases & Information Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Tourism & Hospitality (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Strategic Management (AREA)
  • Development Economics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Educational Administration (AREA)
  • General Business, Economics & Management (AREA)
  • Economics (AREA)
  • Primary Health Care (AREA)
  • Marketing (AREA)
  • Human Resources & Organizations (AREA)
  • Animal Behavior & Ethology (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to a service recommendation method based on deep matrix decomposition and a knowledge graph, which utilizes government service item data to construct a government service knowledge graph, models entity context and entity description text information of the knowledge graph through a knowledge representation method, and combines knowledge representation learning and personalized recommendation through a combined learning mode to obtain an optimal GKGR model. And finally predicting the score of each pair of users and service items, and recommending the service items with higher scores to the users in a recommendation list mode. According to the method, the neural network is utilized to conduct feature extraction on users and service matters, user behavior data is fully utilized, and the problem of data sparseness is effectively relieved; modeling is carried out on the entity context and the entity description text information through a knowledge representation method, knowledge representation tasks and personalized recommendation tasks are jointly learned, accuracy and interpretability of recommendation results are improved, and the problem of cold start of government service recommendation is effectively relieved.

Description

Service recommendation method based on depth matrix decomposition and knowledge graph
Technical Field
The invention relates to the field of government service recommendation, in particular to a service recommendation method based on deep matrix decomposition and knowledge graph.
Background
The Internet and government service combines the traditional government service mode with the modern Internet technology, so that convenient interaction and information communication between the government and citizens are realized, the efficiency, transparency and fairness of the government service are improved, and meanwhile, digital transformation and smart city construction are also promoted. However, with the construction of the urban one-stop service platform, government service resources are huge and dispersed, various and level complex, and the government service is oriented to citizen users, and personalized information service is often required. How to filter out the needed service matters from mass city government service for users and recommend the service matters to the users is a pain and difficulty problem faced by the one-stop city service platform. In the personalized recommendation technology, the traditional collaborative filtering recommendation algorithm is widely applied and mature in technology, but is difficult to deal with the data sparseness problem faced in the personalized recommendation scene of government service.
Disclosure of Invention
The invention aims to provide a service recommendation method based on depth matrix decomposition and a knowledge graph, which aims to solve the problems that user data are sparse, a large amount of heterogeneous multisource and loose data are not fully utilized during government service recommendation.
In order to solve the technical problems, the invention adopts the following technical scheme: a service recommendation method based on depth matrix decomposition and knowledge graph comprises the following steps:
s1, obtaining an initial user vector according to behavior data generated in the interaction process of the user and the service items, and outputting the initial user vector as an input full-connection layer to obtain a user vector u i ：
Quantifying the user behavior data according to the measurement rules according to the behavior data generated in the interaction process of the user and the service items, and constructing a user-service item behavior matrixWherein each row of the matrix represents an initial user vector, the values R in the matrix ij The number of clicks of user i on service item j is indicated.
Constructing a government service knowledge graph G, wherein the service event entities and the relations are represented by graph structures, each service event entity is regarded as a node in the graph, and the relations are regarded as edges;
s2, constructing and training a GKGR model, wherein the GKGR model comprises the following steps:
s2-1, obtaining a service item entity vector e according to the government service knowledge graph definition entity context information s ；
S2-2, obtaining a second service item entity vector e according to the government service knowledge graph definition entity description text d ；
S2-3, according to the service item entity vector e s And service item entity vector e d Obtaining a final service item vector e;
s2-4 given user i, service instance entity j and user-service instance behavior matrix R ij User-service item preference pairs are constructed<i,j,j′>The user i has interaction with the service item entity j and does not have interaction with the service item j', i.e. the user i has a demand for the service item j. The triples and entity description text associated with j, j' are found from G. Learning e by knowledge representation of entity context s Learning service item entity vector e using Bi-LSTM d . The two entity vectors are fused through a gating mechanism, and the user vector u is obtained i And the service item vector is input into the personalized ranking model.
When the objective function is maximum and is not changed any more, training is finished, and an optimal GKGR model is obtained at the moment;
and S3, for a user, adopting S1 to obtain a user vector, inputting an optimal GKGR model, calculating the association degree of the user and all service items by the optimal GKGR model, arranging the service items according to the descending order of the association degree values, and outputting the service item sequence corresponding to the association degree values.
Preferably, the S2-1 obtains a service item entity vector e s The process of (2) is as follows:
the entity context information C (h, r, t) includes neighbor context C n (h) And path context C p (h,t)。
Neighbor context C n (h) Refers to a collection of other nodes that are directly connected to a given node.
Path context C p (h, t) refers to the context information composed of all paths connected to a given node, i.e., C (h, r, t) =c n (h)∪C p (h,t)。
The neighbor context definition of the service item entity h is shown in the following formula, and G represents a government service knowledge graph.
Wherein h, t represent different service event entities, and r represents a relationship;
the path context definition of service instance entities h and t is as follows:
wherein p is i Is the relation sequence of h reaching entity t, L is the maximum length in all relation paths, r 1 ，Representation ofh other relations to the entity t path e 1 ，/>Representing h reaching other entities of the entity t pathway, l i Representing the ith relationship.
The probability that triplet (h, r, t) is established is shown in the following formula.
f(h,r,t)＝P((h,r,t)|C(h,r,t)；θ) (3)
Wherein θ represents a parameter of the model, and the higher the score of the scoring function f (·) is, the greater the probability that the triplet is established.
The pre-training model is transmitted, the triplet (h, r, t) is input into the transmitted, when the f (h, r, t) value is maximum, the output of the transmitted is the service item entity vector e s 。
Preferably, the scoring function f (·) is optimized by:
by decomposing f (h, r, t) with conditional probability, we can get:
f(h,r,t)＝P(h|C(h,r,t)；θ)·P(t|C(h,r,t),h；θ)·P(r|C(h,r,t),h,t；θ) (4)
where P (h|C (h, r, t); θ) represents the conditional probability of occurrence at h. Since entity h is contextually related to its neighbors, P (h|C (h, r, t); θ) can be directly approximated as P (h|C) n (h) The method comprises the steps of carrying out a first treatment on the surface of the θ), defined as the following formula.
Wherein the method comprises the steps ofRepresenting the degree of association of any entity with the h entity neighbor context.
h' represents the head entity in the wrong triplet;
p (t|C (h, r, t), h; θ) represents entity t probability, the degree of association between the head entity and the tail entity is measured by the path context,
p (t|C (h, r, t), h; θ) is approximately expressed as P (t|C) p (h, t), h; θ), defined as the following formula.
Where ε represents the set of tail entities and t' represents the tail entities in the error triplet (negative example);
p (r|C (h, r, t), h, t; θ) represents the conditional probability of the occurrence of the relationship r. Entities h and t have determined that the entity context has been introduced, and therefore, the entity context C (h, r, t) in P (r|C (h, r, t), h, t; θ) is omitted as shown in the following formula.
P (h|C (h, r, t); θ), P (t|C (h, r, t), h; θ) and P (r|C (h, r, t), h, t; θ) in the scoring function are approximately expressed as P (h|C) n (h)；θ)、P(t|C p (g, t), g; θ) and P (r|h, t; θ) as shown in the following formula.
f(g,r,t)≈P(h|C n (h)；θ)·P(t|C p (h, t), h; θ) ·p (r|h, t; θ) (8) by maximizing the scoring function f (h, r, t) =p ((h, r, t) |c (h, r, t); θ) optimizes vectors of entity context information. Preferably, the S2-2 obtains the service item entity vector e according to the government knowledge map entity description text d The process of (2) is as follows:
based on the constructed government service knowledge graph G, defining an entity description text aiming at the service event entity, wherein the entity description text comprises an entity name, an entity association relationship name and a tail entity name;
the weight of the ith position of the entity description text to a given relationship r is defined as α i (r) as shown in the following formula.
Wherein, among them,is a relation vector obtained by expression learning, +.>Is the output of the i-th position, W a And U a Is a parameter matrix,/->Is a parameter vector. e, e i (r) is z i And the relation r, n represents the length of the entity description text.
Service item entity vector e d The definition is shown in the following formula.
x 1 ，x n Representing the positions of the entity description text of length 1 and n, respectively.
Preferably, the S2-3 is based on a service item entity vector e s And service item entity vector e d The process of obtaining the final service event vector e is as follows:
e is controlled by a gating mechanism s And e d And (5) fusing to obtain e, wherein the definition is shown in the following formula.
e＝β⊙e s +(1-β)⊙e d (12)
Wherein β ε [0,1] represents the gate that balances two types of representation weights.
Preferably, the objective function L is:
v j ＝β⊙e sj +(1-β)⊙e dj (15)
wherein u is i Is a vector representation of user i, v j 、v j′ Vector representations of government service event entities j and j', respectively, z representing a canonical term.
Wherein f (h, r, t, d) h ,d t ) Scoring function, g, representing knowledge representation learning portion h 、g t The gating sizes of the head entity and the tail entity, h respectively s 、t s Vector of entity context information h and t, respectively, h d 、t d Vectors of textual knowledge are described for the entities of h and t.
Compared with the prior art, the invention has at least the following advantages:
and S1, constructing a user-service item behavior matrix by analyzing the characteristics of user behaviors, and extracting features of the user and the service items by using a neural network. User behavior data is fully utilized, and the problem of data sparseness is effectively relieved. The problem that the personalized degree of the government service recommendation method is insufficient, and the user-service item matrix is sparse and the like are difficult to solve by the traditional collaborative filtering technology is solved.
And S2, a government service knowledge graph is constructed by using government resource data, modeling is carried out on entity context and entity description text information through a knowledge representation method, and a knowledge representation task and a personalized recommendation task are jointly learned, so that the accuracy and the interpretability of a recommendation result are improved, and the cold start problem of government service recommendation is effectively relieved. The problems that government information resources are heterogeneous and multi-source, and organization is loose and cannot be fully utilized are solved.
Drawings
FIG. 1 illustrates an example of knowledge-based government service recommendations.
Fig. 2 government ontology construction step.
Fig. 3 government body structure.
FIG. 4prot g builds a government body.
FIG. 5 is a partial RDF triplet.
Fig. 6 is a block diagram of an entity description text message.
FIG. 7 is a schematic diagram of the method of the present invention.
FIG. 8 is a diagram of an example triplet.
Detailed Description
The present invention will be described in further detail below.
The invention provides a service recommendation method (BDMF) based on depth matrix decomposition and knowledge graph through modeling user behaviors. And extracting characteristics of the user and the service matters through a neural network, and predicting the demand degree of the user on the non-interactive service matters by utilizing deep collaborative filtering. The BDMF does not consider the impact of service related information on the user's needs. In the government service scenario, each service item includes a plurality of information such as a reception condition, a transaction body, a travel hierarchy, and a service object. The user requirements are closely related to the item information, and the service item information is reasonably utilized, so that the recommendation accuracy of BDMF can be effectively improved. As shown in fig. 1, when the user clicks on the service item "enterprise social security registration", the acceptance condition of the service item is "to be registered in the market regulatory agency", the recommendation algorithm should consider the information to recommend the service item which is implemented by the market regulatory agency and is related to the enterprise to the user, for example: "internal resource enterprises and branches set up registration (company set up registration)", "foreign resource enterprises and branches set up registration (foreign investment enterprises set up registration)", and the like. In government service recommendations, the more similar the service event information is, the higher the reference value to the user's needs. The similarity between service item information is reasonably utilized, the accuracy of an algorithm can be improved, and the problem of cold start of government service recommendation is effectively solved.
In a government service recommendation scenario, a user's potential needs are associated with service event information. However, government service resources are huge and scattered, are multiple in variety and are complex in level, and government service recommendation faces the problems of heterogeneous and multiple sources of government information resources, loose organization and insufficient utilization. The knowledge graph is used as a heterogeneous network containing rich semantic information, the knowledge representation method can be used for effectively representing the entities and the relations in the knowledge graph in a low-dimensional continuous vector space, government resource data can be reasonably utilized, and meanwhile, the capabilities of knowledge graph fusion, reasoning and application are given, so that the accuracy of government service recommendation is improved.
Government service knowledge graph construction
Knowledge maps can be logically divided into a pattern layer and a data layer. The pattern layer is the core of the knowledge graph, stores the refined knowledge, and defines and standardizes the data level and class in the field. An ontology library is generally used for managing a pattern layer of the atlas, and the capabilities of rules, axioms, constraint conditions and the like in the ontology library are utilized for standardizing the association among the entities, relations, types of the entities, attributes and the like in the atlas.
The data layer is responsible for the specific storage of specific triples in the knowledge graph, is structurally under the mode layer, and is an actual expression form of the whole knowledge graph. In the data layer, triples are stored in the graph database in the form of two expressions of < entity, relationship, entity > and < entity, attribute, value >, as shown in fig. 8.
The knowledge graph construction process starts from acquiring the original knowledge data, and a knowledge processing technology (including automatic or semi-automatic) is adopted to extract the required knowledge elements from the original data. And storing according to the definition of the mode layer and the data layer. The massive heterogeneous knowledge forms a huge entity relationship network through the structural definition of a mode layer and the processing of a data layer, so that a knowledge graph is constructed. The knowledge graph is constructed based on a priori knowledge of the domain expert.
A service recommendation method based on depth matrix decomposition and knowledge graph comprises the following steps:
s1, obtaining an initial user vector according to behavior data generated in the interaction process of the user and the service items, and outputting the initial user vector as an input full-connection layer to obtain a user vector u i ：
The user vector is obtained based on the user behavior and the depth matrix decomposition. Quantifying the user behavior data according to the measurement rules according to the behavior data generated in the interaction process of the user and the service items, and constructing a user-service item behavior matrixWherein each row of the matrix represents an initial user vector, the values R in the matrix ij The number of clicks of user i on service item j is indicated.
Constructing a government service knowledge graph G, wherein the service event entities and the relations are represented by graph structures, each service event entity is regarded as a node in the graph, and the relations are regarded as edges;
s2, constructing and training a GKGR model, wherein the GKGR model comprises the following steps:
s2-1, obtaining a service item entity vector e according to the government service knowledge graph definition entity context information s ；
S2-2, obtaining a second service item entity vector e according to the government service knowledge graph definition entity description text d ；
S2-3, according to the service item entity vector e s And service item entity vector e d Obtaining a final service item vector e;
s2-4, learning vector representations of the user, the service item entity and the relationship by combining the user-service item behavior data and the government knowledge graph data by adopting a combined training method. Given user i, service item entity j and user-service item behavior matrix R ij User-service item preference pairs are constructed<i,j,j′>The user i has interaction with the service item entity j and does not have interaction with the service item j', i.e. the user i has a demand for the service item j. The triples and entity description text associated with j, j' are found from G. Learning e by knowledge representation of entity context s Learning service item entity vector e using Bi-LSTM d . And fusing the two entity vectors through a gating mechanism, and inputting the user vector and the service item vector into the personalized sequencing model. PersonalisationThe ordering model adopts Bayesian personalized ordering to convert the recommendation problem into an ordering problem, all articles with recommendation in the system are ordered for each user, articles liked by the user are arranged in front as much as possible, and finally the articles with Top-K in the sequence are recommended to the user.
When the objective function is maximum and is not changed any more, training is finished, and an optimal GKGR model is obtained at the moment;
and S3, for a user, adopting S1 to obtain a user vector, inputting an optimal GKGR model, calculating the association degree of the user and all service items by the optimal GKGR model, arranging the service items according to the descending order of the association degree values, and outputting the service item sequence corresponding to the association degree values.
Specifically, the S2-1 obtains a service item entity vector e s The process of (2) is as follows:
the entity context information includes neighbor contexts and path contexts. In the government service knowledge graph, given a service event entity, its neighbor context nodes include the type of service event, enforcement bodies, exercise hierarchy, and so forth. The knowledge representation method of the introduced entity context encodes the semantic information related to the entity into a high-dimensional vector representation, so that the representation capability of the model is improved. And secondly, the knowledge representation of the entity context is introduced, so that the entity semantic relationship can be better understood, the similarity among entities, the hierarchical structure and the like can be realized, and the interpretability of the recommendation system can be improved.
The entity context information C (h, r, t) includes neighbor context C n (h) And path context C p (h,t)。
Neighbor context C n (h) Refers to a collection of other nodes that are directly connected to a given node.
Path context C p (h, t) refers to the context information composed of all paths connected to a given node, i.e., C (h, r, t) =c n (h)∪C p (h,t)。
The neighbor context definition of the service item entity h is shown in the following formula, and G represents a government service knowledge graph.
Wherein h, t represent different service event entities, and r represents a relationship;
the path context definition of service instance entities h and t is as follows:
wherein p is i Is the relation sequence of h reaching entity t, L is the maximum length in all relation paths, r 1 ，Representing other relationships of h to the entity t pathway, e 1 ，/>Representing the g-arrival at other entities of the entity t-pathway, l i Representing the ith relationship. The probability that triplet (h, r, t) is established is shown in the following formula.
f(h,r,t)＝P((h,r,t)|C(h,r,t)；θ) (3)
Wherein θ represents a parameter of the model, and the higher the score of the scoring function f (·) is, the greater the probability that the triplet is established.
The pre-training model is transmitted, the triplet (h, r, t) is input into the transmitted, when the f (h, r, t) value is maximum, the output of the transmitted is the service item entity vector e s 。
Specifically, the scoring function f (·) is optimized by a method of:
by decomposing f (h, r, t) with conditional probability, we can get:
f(h,r,t)＝P(h|C(h,r,t)；θ)·P(t|C(h,r,t),h；θ)·P(r|C(h,r,t),h,t；θ) (4)
where P (h|C (h, r, t); θ) represents the conditional probability of occurrence at h. Since entity h is contextually related to its neighbors, P (h|C (h, r, t); θ) can be directly approximated as P (h|C) n (h) The method comprises the steps of carrying out a first treatment on the surface of the θ), defined as the following formula.
Wherein the method comprises the steps ofRepresenting the degree of association of any entity with the h entity neighbor context.
h' represents the head entity in the wrong triplet (i.e., negative example);
p (t|C (h, r, t), h; θ) represents entity t probability, the degree of association between the head entity and the tail entity is measured by the path context,
p (t|C (h, r, t), h; θ) is approximately expressed as P (t|C) p (h, t), h; θ), defined as the following formula.
Where ε represents the set of tail entities and t' represents the tail entities in the error triplet (negative example).
P (r|C (h, r, t), h, t; θ) represents the conditional probability of the occurrence of the relationship r. Entities h and t have determined that the entity context has been introduced, and therefore, the entity context C (h, r, t) in P (r|C (h, r, t), h, t; θ) is omitted as shown in the following formula.
P (h|C (h, r, t); θ), P (t|C (h, r, t), h; θ) and P (r|C (h, r, t), h, t; θ) in the scoring function are approximately expressed as P (h|C) n (h)；θ)、P(t|C p (h, t), h; θ) and P (r|h, t; θ) as shown in the following formula.
f(h,r,t)≈P(h|C n (h)；θ)·P(t|C p (h,t),h；θ)·P(r|h,t；θ) (8)
The vector of entity context information is optimized by maximizing the scoring function f (h, r, t) =p ((h, r, t) |c (h, r, t); θ).
Specifically, the S2-2 obtains a service item entity vector e according to the government knowledge map entity description text d The process of (2) is as follows:
based on the constructed government service knowledge graph G, defining an entity description text aiming at the service event entity, wherein the entity description text comprises an entity name, an entity association relationship name and a tail entity name;
the weight of the ith position of the entity description text to a given relationship r is defined as α i (r) as shown in the following formula.
Wherein, among them,is a relation vector obtained by expression learning, +.>Is the output of the i-th position, W a And U a Is a parameter matrix,/->Is a parameter vector. e, e i (r)z i And the relation r, n represents the length of the entity description text.
Service item entity vector e d The definition is shown in the following formula.
x 1 ，x n Representing the positions of the entity description text of length 1 and n, respectively.
Specifically, the S2-3 is according toService item entity vector e s And service item entity vector e d The process of obtaining the final service event vector e is as follows:
e is controlled by a gating mechanism s And e d And (5) fusing to obtain e, wherein the definition is shown in the following formula.
e＝β⊙e s +(1-β)⊙e d (12)
Wherein β ε [0,1] represents the gate that balances two types of representation weights.
Specifically, the objective function L is:
v j ＝β⊙e sj +(1-β)⊙e dj (15)
wherein u is i Is a vector representation of user i, v j 、v j′ Vector representations of government service event entities j and j', respectively, z representing a canonical term.
Wherein f (h, r, t, d) h ,d t ) Scoring function, g, representing knowledge representation learning portion h 、g t The gating sizes of the head entity and the tail entity, h respectively s 、t s Vector of entity context information h and t, respectively, h d 、t d Vectors of textual knowledge are described for the entities of h and t. h is a s 、r、t s Is obtained by pretraining based on a TransE representation learning method, h d 、t d Is obtained by performing representation learning on entity description text.
TABLE 1 GKGR training procedure
Experimental design and analysis
According to the invention, a service recommendation method based on user behavior and depth matrix decomposition is used as an experimental baseline, the influence on a service recommendation result after a knowledge representation module is added is analyzed, and a recommendation method for merging government service knowledge graphs is verified.
1. Data set
The construction mode of the negative example triplet is as follows: and randomly selecting a service item which is never interacted by a user from the user-service item behavior data as a negative example, searching a triplet containing the service item entity in a government service knowledge graph, and carrying out negative sampling on the triplet. The negative sampling mode is to replace the head entity or the tail entity in the triples to obtain the triples and entity description text which do not exist in the knowledge graph. By increasing the diversity of negative examples, the algorithm can learn the relation between the positive examples of triples better, and the recommendation accuracy is improved. Finally, the sampled negative examples and positive examples are used for training and testing algorithms together, and the positive and negative proportions are 1:3.
2. evaluation index
In order to ensure the consistency of the experiment, the evaluation index is the same as the index used in the service recommendation method experiment based on the user behavior and the depth matrix decomposition, namely HR, precision, recall and F1 values.
Each model vector dimension in the experiment was set to 64 dimensions. The network parameters were initialized using a gaussian distribution with a mean of 0 and a variance of 0.001. The batch size is set to 64, the positive and negative sample ratio is 1:3, the learning rate is set to 0.01, the regularization coefficients are all 0.001, and an Adam optimizer is adopted for parameter optimization.
BDMF: the method provided by the invention.
BDMF+TransE: the entity is represented using a TransE model.
CKE: recommendation system methods based on collaborative knowledge base embedding.
(1) Comparing the difference between the recommended effect of the method and the comparison method
The experimental results of each model under the HR, precision, recall and F1 indexes are shown in Table 2.
Table 2 comparative experimental results
According to experimental result analysis, the GKGR, CKE, BDMF +TransE model is higher than the BDMF method in four indexes after the government service knowledge maps are fused, wherein the GKGR model provided by the invention is respectively improved by 5.08%,17.53%,10.58% and 22.87% in Precision, recall, F and HR. In the model integrating the government service knowledge graph, the BDMF+TransE method is lower than the GKGR model and the CKE model in the related indexes because the entity description text is not modeled. The CKE model introduces entity description text on the basis of representing the government service knowledge graph through the TransE, so that the recommendation effect is improved to a certain extent. The GKGR model provided by the invention realizes representation of the entity path context and the neighbor context through knowledge representation of the entity context on the basis of CKE, and compared with CKE, precision, recall, F and HR, the GKGR model respectively improves 5.76%,3.77%,5.6% and 9.95%, improves the effect of government service recommendation, and shows that the performance can be further improved by fusing the government service knowledge graph entity context.
In the government service recommendation scenario, if the path context and the neighbor context of two government service event entities are similar in the knowledge graph, the entity vector representations will be correspondingly closer. When a government service platform handles a service, a user usually finds a front service of the service item according to information such as a reception condition and a travel level of the service item. By describing text characterization for service item entities, semantic representation capability of the knowledge graph is enhanced. The recommendation model fused with the government service knowledge graph fully utilizes the relationship among government service event entities, introduces the entity context and entity description text of the government service knowledge graph, balances the weight occupied by the entity description text and the entity context information through a gating mechanism, and can realize better recommendation effect.
(2) Comparing the influence of different knowledge representation methods on the recommendation effect
In order to verify the influence of the knowledge representation method on the recommendation result, four common TransX series models, namely TransE, transH, transR and TransD, are selected. And measured using Precision, recall, F, HR evaluation index. The recommended effect after the method was expressed using different knowledge as shown in table 3. Experimental results show that TransH, transR, transD performs better than TransE in government service recommendation tasks, wherein TransD knowledge shows that the method performs best on two indexes of recommendation accuracy and recall rate, the F1 value is 0.2897, and the HR value is 0.7661. However, the effect of the TransE is less different from that of the TransD, and the model is simple in structure and easy to train, so that the experiment is mainly trained based on the TransE model.
TABLE 3 influence of different representation methods on respective indicators
(3) Comparing differences of different models on recommending effect of entity description text
In order to verify the effect of the entity description text on the recommendation result, four common models are selected for representing the entity description text, namely Word2Vec, CNN, RNN and Bi-LSTM, and are measured by using Precision, recall, F and HR evaluation indexes. As shown in table 4, the recommended effect on the different characterization methods was used.
TABLE 4 influence of different characterization methods on the respective indicators
As shown in Table 4, on modeling of entity description text, RNN and Bi-LSTM which are good at processing text sequences have better effect, and have stronger extraction capability on entity description text characteristics. The Bi-LSTM introduces a gating mechanism on the basis of RNN, has stronger modeling capability on entity description text, is 6.5 percent and 9.25 percent higher than RNN on F1 and HR indexes respectively, and improves the accuracy of government service recommendation.
Finally, it is noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made thereto without departing from the spirit and scope of the technical solution of the present invention, which is intended to be covered by the scope of the claims of the present invention.

Claims (6)

1. A service recommendation method based on depth matrix decomposition and knowledge graph is characterized by comprising the following steps:
s1: obtaining an initial user vector according to behavior data generated in the interaction process of the user and the service items, and outputting the initial user vector as an input full-connection layer to obtain a user vector u i ：
Quantifying the user behavior data according to the measurement rules according to the behavior data generated in the interaction process of the user and the service items, and constructing a user-service item behavior matrixWherein each row of the matrix represents an initial user vector, the values R in the matrix ij The number of clicks of the user i on the service item j is represented;
constructing a government service knowledge graph G, wherein the service event entities and the relations are represented by graph structures, each service event entity is regarded as a node in the graph, and the relations are regarded as edges;
s2: constructing and training a GKGR model, said GKGR model comprising:
s2-1: obtaining service item entity vector e according to government service knowledge graph definition entity context information s ；
S2-2: obtaining a second service item entity vector e according to the government service knowledge graph definition entity description text d ；
S2-3: entity according to service mattersVector e s And service item entity vector e d Obtaining a final service item vector e;
s2-4: given user i, service item entity j and user-service item behavior matrix R ij User-service item preference pairs are constructed<i，j，j′>Indicating that the user i has interaction with the service item entity j and does not have interaction with the service item j ', namely that the user i has a requirement on the service item j, finding out triples and entity description text related to j and j' from G, and learning e by a knowledge representation method of entity context s Learning service item entity vector e using Bi-LSTM d The two entity vectors are fused through a gating mechanism, and the user vector u is obtained i And the service item vector is input into the personalized sequencing model;
when the objective function is maximum and is not changed any more, training is finished, and an optimal GKGR model is obtained at the moment;
s3: for a user, S1 is adopted to obtain a user vector and input an optimal GKGR model, the optimal GKGR model calculates the association degree of the user and all service matters, and the service matters are arranged in descending order according to the association degree value, and the service matter sequence corresponding to the association degree value is output.
2. The service recommendation method based on depth matrix decomposition and knowledge graph as claimed in claim 1, wherein: the S2-1 obtains a service item entity vector e s The process of (2) is as follows:
the entity context information C (h, r, t) includes neighbor context C n (h) And path context C p (h，t)；
Neighbor context C n (h) Refers to a collection of other nodes directly connected to a given node;
path context C p (h, t) refers to the context information composed of all paths connected to a given node, i.e., C (h, r, t) =c n (h)∪C p (h，t)；
The neighbor context definition of the service item entity h is shown in the following formula, and G represents a government service knowledge graph;
wherein h, t represent different service event entities, and r represents a relationship;
the path context definition of service instance entities h and t is as follows:
wherein p is i Is the relation sequence of h reaching entity t, L is the maximum length in all relation paths, r 1 ，Representing other relationships of h to the entity t pathway, e 1 ，/>Representing h reaching other entities of the entity t pathway, l i Representing an ith relationship;
the probability that triplet (h, r, t) is established is shown in the following formula;
f(h，r，t)＝P((h，r，t)|C(h，r，t)；θ) (3)
wherein θ represents a parameter of the model, and the higher the score of the scoring function f (·) is, the greater the probability of the triplet being established;
the pre-training model is transmitted, the triplet (h, r, t) is input into the transmitted, when the f (h, r, t) value is maximum, the output of the transmitted is the service item entity vector e s 。
3. The service recommendation method based on depth matrix decomposition and knowledge graph as claimed in claim 2, wherein: optimizing the scoring function f (·) by adopting a method:
by decomposing f (h, r, t) with conditional probability, we can get:
f(h，r，t)＝P(h|C(h，r，t)；θ)·P(t|C(h，r，t)，h；θ)·P(r|C(h，r，t)，h，t；θ) (4)
wherein P (h|C (h, r, t); θ) represents the conditional probability of occurrence at h, defined as shown in the following formula;
wherein the method comprises the steps ofRepresenting the association degree of any entity and a h entity neighbor context;
h' represents the head entity in the wrong triplet;
p (t|C (h, r, t), h; θ) represents entity t probability, the degree of association between head and tail entities is measured by path context, and P (t|C (h, r, t), h; θ) is approximately represented as P (t|C) p (h, t), h; θ), defined as the following formula;
where ε represents the set of tail entities and t' represents the tail entities in the error triplet (negative example);
p (r|C (h, r, t), h, t; θ) represents the conditional probability of occurrence of the relationship r; entities h and t have determined that the entity context has been introduced, and therefore omitting the entity context C (h, r, t) in P (r|C (h, r, t), h, t; θ) is shown in the following formula;
p (h|C (h, r, t); θ), P (t|C (h, r, t), h; θ) and P (r|C (h, r, t), h, t; θ) in the scoring function are approximately expressed as P (h|C) n (h)；θ)、P(t|C p (h, t), h; θ) and P (r|h, t; θ) as shown in the following formula;
f(h，r，t)≈P(h|C n (h)；θ)·P(t|C p (h，t)，h；θ)·P(r|h，t；θ) (8)
the vector of entity context information is optimized by maximizing the scoring function f (h, r, t) =p ((h, r, t) |c (h, r, t); θ).
4. The service recommendation method based on depth matrix decomposition and knowledge graph as claimed in claim 3, wherein: s2-2 obtains a service item entity vector e according to the government knowledge map entity description text d The process of (2) is as follows:
based on the constructed government service knowledge graph G, defining an entity description text aiming at the service event entity, wherein the entity description text comprises an entity name, an entity association relationship name and a tail entity name;
the weight of the ith position of the entity description text to a given relationship r is defined as α i (r) as shown in the following formula;
wherein, among them,is a relation vector obtained by expression learning, +.>Is the output of the i-th position, W a And U a Is a parameter matrix,/->Is a parameter vector; e, e i (r) is z i Closing deviceThe correlation of r, n represents the length of the entity description text;
service item entity vector e d The definition is shown in the following formula;
x 1 ，x n representing the positions of the entity description text of length 1 and n, respectively.
5. The service recommendation method based on depth matrix decomposition and knowledge graph as claimed in claim 4, wherein: the S2-3 is based on the service item entity vector e s And service item entity vector e d The process of obtaining the final service event vector e is as follows:
e is controlled by a gating mechanism s And e d Fusing to obtain e, wherein the definition is shown in the following formula;
e＝β⊙e s +(1-β)⊙e d (12)
wherein β ε [0,1] represents the gate that balances two types of representation weights.
6. The service recommendation method based on depth matrix decomposition and knowledge graph as claimed in claim 5, wherein: the objective function L is:
v j ＝β⊙e sj +(1-β)⊙e dj (15)
wherein u is i Is a vector representation of user i, v j 、v j′ Vector table of government service event entities j and j' respectivelyShown, z represents a regularization term;
wherein f (h, r, t, d) h ，d t ) Scoring function, g, representing knowledge representation learning portion h 、g t The gating sizes of the head entity and the tail entity, h respectively s 、t s Vector of entity context information h and t, respectively, h d 、t d Vectors of textual knowledge are described for the entities of h and t.
CN202310584709.XA 2023-05-23 2023-05-23 Service recommendation method based on depth matrix decomposition and knowledge graph Pending CN116662564A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310584709.XA CN116662564A (en) 2023-05-23 2023-05-23 Service recommendation method based on depth matrix decomposition and knowledge graph

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310584709.XA CN116662564A (en) 2023-05-23 2023-05-23 Service recommendation method based on depth matrix decomposition and knowledge graph

Publications (1)

Publication Number Publication Date
CN116662564A true CN116662564A (en) 2023-08-29

Family

ID=87727202

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310584709.XA Pending CN116662564A (en) 2023-05-23 2023-05-23 Service recommendation method based on depth matrix decomposition and knowledge graph

Country Status (1)

Country Link
CN (1) CN116662564A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117633540A (en) * 2024-01-25 2024-03-01 杭州阿里云飞天信息技术有限公司 Sample data construction method and device
CN118193842A (en) * 2024-04-02 2024-06-14 北京绿能碳宝科技发展有限公司 Interpretable recommendation method and system based on causal reasoning

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117633540A (en) * 2024-01-25 2024-03-01 杭州阿里云飞天信息技术有限公司 Sample data construction method and device
CN117633540B (en) * 2024-01-25 2024-04-30 杭州阿里云飞天信息技术有限公司 Sample data construction method and device
CN118193842A (en) * 2024-04-02 2024-06-14 北京绿能碳宝科技发展有限公司 Interpretable recommendation method and system based on causal reasoning

Similar Documents

Publication Publication Date Title
CN113158033A (en) Collaborative recommendation model construction method based on knowledge graph preference propagation
CN113779264B (en) Transaction recommendation method based on patent supply and demand knowledge graph
CN112966091B (en) Knowledge map recommendation system fusing entity information and heat
CN116662564A (en) Service recommendation method based on depth matrix decomposition and knowledge graph
WO2023179689A1 (en) Knowledge graph-based recommendation method for internet of things
CN113590965B (en) Video recommendation method integrating knowledge graph and emotion analysis
Özdağoğlu et al. A predictive filtering approach for clarifying bibliometric datasets: an example on the research articles related to industry 4.0
Okawa et al. Predicting opinion dynamics via sociologically-informed neural networks
CN113918833A (en) Product recommendation method realized through graph convolution collaborative filtering of social network relationship
CN111723285A (en) Depth spectrum convolution collaborative filtering recommendation method based on scores
Li et al. Capsule neural tensor networks with multi-aspect information for Few-shot Knowledge Graph Completion
CN116340524B (en) Method for supplementing small sample temporal knowledge graph based on relational adaptive network
CN115840853A (en) Course recommendation system based on knowledge graph and attention network
Yu et al. A graph attention network under probabilistic linguistic environment based on Bi-LSTM applied to film classification
CN114564594A (en) Knowledge graph user preference entity recall method based on double-tower model
CN112784049B (en) Text data-oriented online social platform multi-element knowledge acquisition method
CN110299194A (en) The similar case recommended method with the wide depth model of improvement is indicated based on comprehensive characteristics
Qin et al. Towards a personalized movie recommendation system: A deep learning approach
CN116452241B (en) User loss probability calculation method based on multi-mode fusion neural network
Kobren et al. Constructing high precision knowledge bases with subjective and factual attributes
CN117892815A (en) Graph comparison recommendation method based on knowledge graph
CN117056609A (en) Session recommendation method based on multi-layer aggregation enhanced contrast learning
CN117196716A (en) Digital signage advertisement theme recommendation method based on Transformer network model
Li et al. Graph structure estimation neural network-based service classification
CN113190733B (en) Network event popularity prediction method and system based on multiple platforms

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination