CN110689423B - Credit evaluation method and device - Google Patents
Credit evaluation method and device Download PDFInfo
- Publication number
- CN110689423B CN110689423B CN201910780782.8A CN201910780782A CN110689423B CN 110689423 B CN110689423 B CN 110689423B CN 201910780782 A CN201910780782 A CN 201910780782A CN 110689423 B CN110689423 B CN 110689423B
- Authority
- CN
- China
- Prior art keywords
- user
- credit
- target user
- track
- target
- 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.)
- Active
Links
Abstract
The invention discloses a credit evaluation method and a credit evaluation device, relates to the technical field of data processing, and aims to solve the problem that credit scores calculated in the prior art cannot accurately evaluate personal financial rights of clients. The method mainly comprises the following steps: acquiring user track labels according to location-based service (LBS) data of each system user; constructing a knowledge graph according to the user interest labels, the user track labels and the user call records of each system user, wherein the knowledge graph is used for identifying the association relation among all the system users; extracting user feature vectors by using a Node2vec algorithm according to the knowledge graph; inputting user information of a target user into a preset neural network model, and calculating credit probability of whether the target user breaks a contract or not; and inputting the credit probability into a preset scoring card model, and calculating the credit score of the target user. The invention is mainly used in the financial authority assessment process.
Description
Technical Field
The present invention relates to the field of data processing technologies, and in particular, to a method and apparatus for credit evaluation.
Background
Credit evaluation refers to evaluating personal financial rights based on the credit history of the customer. In the credit evaluation process, a certain credit score model is utilized to obtain credit scores of different grades, and then a creditor analyzes the possibility of timely repayment of a client according to the credit scores to determine whether to grant credit and credit limit and interest rate of the credit, so that the safety of a financial department is ensured.
In the prior art, at least one category of customer information of a sample customer is obtained, then customer information of each category is preprocessed to obtain sampling data, then modeling is conducted according to the sampling data corresponding to the customer information of each category and the default information of the sample customer to generate a default risk scoring model corresponding to the customer information of each category, then the default risk scoring model corresponding to the customer information of each category is subjected to model fusion to obtain a credit scoring model, and finally the customer information of the customer to be evaluated is input into the credit scoring model to calculate the credit score of the customer to be evaluated.
In the method in the prior art, the client information of the sample client is mutually independent, the correlation between the data and the default information is weak, the credit score obtained by calculation of the credit score model obtained by combining the default risk score models corresponding to the client information of different categories has limited influence on whether the sample client is default, and the calculated credit score cannot accurately evaluate the personal financial rights of the client.
Disclosure of Invention
In view of the above, the present invention provides a credit evaluation method and apparatus, and is mainly aimed at solving the problem that credit scores calculated in the prior art cannot accurately evaluate personal financial rights of customers.
According to one aspect of the present invention, there is provided a method of credit assessment, comprising:
acquiring user track labels according to location-based service (LBS) data of each system user;
constructing a knowledge graph according to the user interest labels, the user track labels and the user call records of each system user, wherein the knowledge graph is used for identifying the association relation among all the system users;
extracting user feature vectors by using a Node2vec algorithm according to the knowledge graph;
inputting user information of a target user into a preset neural network model, and calculating credit probability of whether the target user breaks down, wherein the user information comprises user personal information of the target user, user interest labels of the target user, user track labels of the target user, user feature vectors and user call records of the target user;
and inputting the credit probability into a preset scoring card model, and calculating the credit score of the target user.
According to another aspect of the present invention, there is provided an apparatus for credit assessment, comprising:
the acquisition module is used for acquiring user track labels according to the LBS data of each system user based on the location service;
the building module is used for building a knowledge graph according to the user interest labels, the user track labels and the user call records of each system user, and the knowledge graph is used for identifying the association relation among all the system users;
the extraction module is used for extracting user feature vectors by using a Node2vec algorithm according to the knowledge graph;
the calculation module is used for inputting user information of a target user into a preset neural network model, and calculating credit probability of whether the target user breaks down, wherein the user information comprises user personal information of the target user, user interest labels of the target user, user track labels of the target user, user feature vectors and user call records of the target user;
the calculation module is further used for inputting the credit probability into a preset scoring card model and calculating the credit score of the target user.
According to still another aspect of the present invention, there is provided a storage medium having stored therein at least one executable instruction for causing a processor to perform operations corresponding to the method of credit assessment described above.
According to still another aspect of the present invention, there is provided a computer apparatus including: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;
the memory is used for storing at least one executable instruction, and the executable instruction enables the processor to execute the operation corresponding to the credit evaluation method.
By means of the technical scheme, the technical scheme provided by the embodiment of the invention has at least the following advantages:
the invention provides a credit evaluation method and a credit evaluation device, which are characterized in that firstly, user track labels of users of each system are obtained, then, a knowledge graph is constructed according to user interest labels, user track labels and user call records of the users of each system, then, user feature vectors are extracted according to the knowledge graph and a Node2vec algorithm, user information of a target user is input into a preset neural network model, whether the credit probability of the target user violating a contract is calculated, and finally, the credit probability is input into a preset scoring card model to calculate the credit score of the target user. Compared with the prior art, the method and the device have the advantages that through the knowledge graph constructed by the user track labels, the user interest labels and the user call records, the association relation between the users is extracted for credit evaluation, the credit of the individuals is evaluated in the partition group, the accuracy of evaluation data for credit evaluation is improved, and the effectiveness of the credit score obtained through calculation is improved. And extracting the user feature vector in the knowledge graph by using a Node2vec algorithm, so that the feature accuracy of the user feature vector reaction can be increased. The credit probability of the user is predicted through the neural network model, so that the credit score is more accurate, and the personal financial rights of the user such as the credit line, the credit card passing rate, the application amount and the like can be accurately estimated through the credit score.
The foregoing description is only an overview of the present invention, and is intended to be implemented in accordance with the teachings of the present invention in order that the same may be more clearly understood and to make the same and other objects, features and advantages of the present invention more readily apparent.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to designate like parts throughout the figures. In the drawings:
FIG. 1 is a flow chart of a method for credit assessment provided by an embodiment of the invention;
FIG. 2 is a flow chart of another method for credit assessment provided by an embodiment of the invention;
FIG. 3 is a block diagram showing the components of a credit assessment apparatus according to an embodiment of the present invention;
FIG. 4 is a block diagram showing another apparatus for credit assessment according to an embodiment of the invention;
fig. 5 shows a schematic structural diagram of a computer device according to an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
With technological development and technological progress, more and more financial services can be handled through a network. In the processing of financial transactions, especially loan transactions, over a network, it is important to set personal financial rights for network users. The personal financial rights include: the amount of money paid, the credit card passing rate, the application amount, etc. The financial business is processed through the network, an APP is required to be used as a platform, a user becomes a system user in a registration mode, user personal information of the system user is added and generated in the registration and use processes, a server maintaining APP operation carries out credit evaluation on the system user according to the personal information of the system user, and then personal financial rights of the system user are determined according to credit evaluation results. The embodiment of the invention provides a credit evaluation method, as shown in fig. 1, which comprises the following steps:
101. And acquiring user track labels according to the LBS data of each system user based on the location service.
User trajectory tags refer to user interests such as golf, musical instruments, tea, western style food, and the like. LBS data, which is geographical location information, is geographical location information of a system user monitored by an APP applying the scheme of the present invention, and is generally represented by longitude and latitude. By means of the LBS data, it is possible to locate the place where the user of the system arrives, such as golf course, string, tea building, western restaurant, etc., and to convert the place into corresponding user track label. In the present invention, it is necessary to acquire user track labels of all system users that have been registered. In the process of obtaining, the system users are ordered, and the user track labels of each system user are obtained according to the order, wherein the ordering basis can be user name, registration time, online time accumulation or loan amount.
102. And constructing a knowledge graph according to the user interest labels, the user track labels and the user call records of each system user.
The user interest tag refers to the interest of the system user, and is obtained according to the historical behavior data of the system user, wherein the historical behavior data comprises self operation, query data, browsing data and the like. For example, when a system user purchases articles such as a badminton racket and a badminton, the user interest tag of the system user indicates that the badminton is the user interest tag of the system user, and when the system user frequently browses financial products, the user interest tag of the system user indicates that the financial is the system user.
The knowledge graph is used for identifying the association relation among all the system users. And in the knowledge graph, system users are taken as nodes, and if two different system users have the same user interest labels, user track labels or user call records, the connection relationship between the two different system users is established. The method for representing the association relationship between different system users can establish a mapping relationship table for each system user, wherein the mapping relationship table comprises the system users with association relationship with each system user, and identifiers can be added for the system users, and the identifiers can identify unique system users, namely the system users with the same identifiers and have association relationship with the same system user.
103. And extracting the user feature vector by using a Node2vec algorithm according to the knowledge graph.
The Node2vec algorithm is a graph embedding method which comprehensively considers depth-first search DFS neighborhood and breadth-first search BFS neighborhood. In the process of extracting the user feature vector, the repayment information of the system users needs to be obtained, the potential association relation among the system users in the knowledge graph is used as the basis, the possible repayment conditions of each system user are calculated, and the repayment conditions are used as the user feature vector values.
104. And inputting user information of the target user into a preset neural network model, and calculating the credit probability of whether the target user breaks the contract.
The user information comprises user personal information of the target user, user interest labels of the target user, user track labels of the target user, user feature vectors and user call records of the target user. All the information in the user information is represented by a single row vector, the user information is spliced before the user information is input into a preset neural network model, and the spliced user information is also a single row vector.
The preset neural network model can be a fixed model, or can be selected according to the loan state of the target user. The loan status includes three states of pre-loan, mid-loan and post-loan. The preset neural network model can be a deep FM network model, an NFM network model and an AFM network model. Before the preset neural network model is used, the preset neural network model needs to be trained according to training data, and the trained preset neural network model can be used.
Credit probability refers to the probability that the target user does not violate. In fact, during the APP operation process, the user feature vector is calculated by continuously accumulating information. When a user of a certain system performs registration, consumption, loan and other operations, if the credit probability needs to be calculated, the credit probability is calculated by taking the user of the system as a target user.
105. And inputting the credit probability into a preset scoring card model, and calculating the credit score of the target user.
A scoring card model is preset for converting the credit probability into a credit score, and the formula can be s=a-B log ((1-p)/p), where S is the credit score, a and B are constants, and p is the credit probability. In the calculation process, two assumptions are required to be set, an expected score P is set for a preset ratio x, a multiple score PDO corresponding to the preset ratio 2x twice is determined, and the multiple score PDO is substituted into a formula of a preset score card model to calculate values of constants A and B, wherein the P and the PDO can be any value between positive infinity and negative infinity. And calculating credit scores corresponding to the credit probabilities of the target users according to the preset score card model, and judging whether the target users can loan, loan amount or repayment strategy according to the credit scores.
The invention provides a credit evaluation method, which comprises the steps of firstly obtaining a user track label of each system user, then constructing a knowledge graph according to a user interest label, a user track label and a user call record of each system user, extracting a user feature vector according to the knowledge graph and a Node2vec algorithm, inputting user information of a target user into a preset neural network model, calculating the credit probability of whether the target user violates, and finally inputting the credit probability into a preset scoring card model to calculate the credit score of the target user. Compared with the prior art, the method and the device have the advantages that through the knowledge graph constructed by the user track labels, the user interest labels and the user call records, the association relation between the users is extracted for credit evaluation, the credit of the individuals is evaluated in the partition group, the accuracy of evaluation data for credit evaluation is improved, and the effectiveness of the credit score obtained through calculation is improved. And extracting the user feature vector in the knowledge graph by using a Node2vec algorithm, so that the feature accuracy of the user feature vector reaction can be increased. The credit probability of the user is predicted through the neural network model, so that the credit score is more accurate, and the personal financial rights of the user such as the credit line, the credit card passing rate, the application amount and the like can be accurately estimated through the credit score.
An embodiment of the present invention provides another credit evaluation method, as shown in fig. 2, including:
201. and acquiring user track labels according to the LBS data of each system user based on the location service.
The user track label refers to a place which is reached by a system user and is counted to obtain an interesting object of the user. The method specifically comprises the following steps: according to preset frequency, sequentially collecting LBS data in each system user, wherein the LBS data are represented by longitude and latitude coordinates; in the electronic map, taking the LBS data as a coordinate center point to obtain map marking places within a preset distance range; and determining the map marking points as user track labels corresponding to the system users.
In order to give consideration to data acquisition and data processing pressure, LBS data of each system user are acquired according to preset frequency. At the client APP, LBS data is collected in real time, and the server collects LBS data of the system user according to a preset frequency. The electronic map may be a different data source, such as a Goldmap, a Baidu map or a Tencel map, and in the embodiment of the present invention, the data source of the electronic map is not limited. Because the terminal of the system user may have deviation during positioning, the coordinate center point corresponding to the distance LBS data is a map mark point in a preset distance range, and is determined as the user track label of the system user. Map marking points refer to places which can be identified on an electronic map, such as national tax authorities, golf courses, certain fitness centers and the like.
202. And constructing a knowledge graph according to the user interest labels, the user track labels and the user call records of each system user.
The knowledge graph is used for identifying the association relation among all the system users. Constructing a knowledge graph, which specifically comprises the following steps: searching the associated users which are the same as the user interest labels, the user track labels and the user call records of each system user; establishing a mapping relation between each system user and the associated user; and constructing the knowledge graph according to the mapping relation.
The graphic identifier for the knowledge graph can intuitively see the association relationship among the users of the system, but the knowledge graph can be stored in a data table form for the convenience of storage and use. The mapping relationship between the system user and the associated user can be the same association relationship which indicates that the system user and the associated user have correlation, or the interest tag association relationship, the track tag association relationship and the call record association relationship are set according to different association contents. And then constructing a knowledge graph according to the mapping relation.
203. And extracting the user feature vector by using a Node2vec algorithm according to the knowledge graph.
In the same APP, only one user feature vector can be extracted at the same time. Extracting a user feature vector, specifically including: sequentially taking each system user as a target user, and calculating the default probability of the target user by using a Node2vec algorithm; and collecting the default probabilities and combining to generate the user feature vector. The user feature vector is a set of the default probabilities of all system users, the user feature vector is a row vector, and the element values in the user feature vector are the default probabilities of the system users. And sequencing the system users according to a preset sequencing rule, sequentially taking each system user as a target user, and calculating the default probability of the target user. The preset ordering rules may be arranged according to the first letter sequence of the names of the system users, may be arranged according to the registration time of the system users, and may be arranged according to the accumulated online time length of the system users, and in the embodiment of the present invention, the specific method of the preset ordering rules is not limited.
Calculating the default probability of the target user, which specifically comprises the following steps: in the knowledge graph, according to a random walk algorithm, obtaining a preset number of node users corresponding to the target users, wherein the node users and the target users have an association relationship, the association relationship comprises a first-level association relationship directly connected between the target users and the node users, and a second-level association relationship indirectly connected between the target users and the node users; searching whether the loan user in the node users violates the constraint information; and calculating the default probability of the target user according to the default information and the association weight corresponding to the association relation.
204. And inputting user information of the target user into a preset neural network model, and calculating the credit probability of whether the target user breaks the contract.
The user information comprises user personal information of the target user, user interest labels of the target user, user track labels of the target user, user characteristic vectors and user call records of the target user. To be able to calculate a more accurate credit probability, more rich information may be added to the user information. The user information also includes a tag frequency and a tag correlation coefficient of a user track tag of the target user, and/or a social track tag of the target user, a tag frequency and a tag correlation coefficient of a social track tag. If in the user informationThe method comprises the steps of calculating corresponding user information before calculating credit probability, wherein the corresponding user information comprises label frequency and label correlation coefficient of a user track label, and/or label frequency and label correlation coefficient of a social track label and a social track label, and the specific steps include: in a preset time range, counting the label frequency of the user track label of the target user, and calculating the label correlation coefficient of the user track label according to a Pearson correlation coefficient formula, wherein the Pearson correlation coefficient formula is that Wherein ρ is X,Y Is the label correlation coefficient of the user track label, X is the user track label, Y is the system user default information with the user track label, cov (X, Y) is the covariance between X and Y, sigma X Is the standard deviation of X, sigma Y Is the standard deviation of Y; and/or, according to the authorization information of the target user, acquiring a social track label corresponding to social LBS data reported by a social media APP, counting the label frequency of the social track label, and calculating the label correlation coefficient of the social track label according to the Pearson correlation coefficient formula, wherein the Pearson correlation coefficient formula is->Wherein ρ is X,Y Is the tag correlation coefficient of the social track tag, X is the social track tag, Y is the information whether the system user with the social track tag violates the constraint, cov (X, Y) is the covariance between X and Y, sigma X Is the standard deviation of X, sigma Y Is the standard deviation of Y.
The label frequency value of the user track label can be selected, the frequency of the user track label of the TOP10 of the last three days of the target user, the frequency of the user track label of the TOP100 of the last week of the target user, the frequency of the user track label of the TOP100 of the last month of the target user, and the label correlation coefficient is the same as the statistical range of the label frequency value. Illustratively, there are 4 system users A, B, C, D with golf user trajectory tags in the current loan system, corresponding to whether to repayment or not 1,0, 1, wherein 1 indicates a violation and 0 indicates no violation, each user trajectory tag has a different value, and the value of the user trajectory tag (golf) is assumed to be 10, and then the tag correlation coefficients are calculated in a matrixFor input, a tag correlation coefficient of the golf user trajectory tag is calculated. The number of the label correlation coefficients corresponds to the number of the user track labels one by one.
Social media APP refers to applications including real-time information interaction such as instant chat APP, news push APP, map APP and the like. Social LBS data is not reported on a fixed frequency, but rather is reported when a user accesses the social media APP. The frequency of the social track labels of the TOP10 can be selected, the frequency of the social track labels of the TOP10 of the last three days of the target user, the frequency of the social track labels of the TOP10 of the last week of the target user, the frequency of the social track labels of the TOP10 of the last month of the target user are the same as the statistical range of the label correlation coefficient.
205. And inputting the credit probability into a preset scoring card model, and calculating the credit score of the target user.
A scoring card model is preset for converting the credit probability into a credit score, and the formula can be s=a-B log ((1-p)/p), where S is the credit score, a and B are constants, and p is the credit probability. In the calculation process, two assumptions are required to be set, an expected score P is set for a preset ratio x, a multiple score PDO corresponding to the preset ratio 2x twice is determined, and the multiple score PDO is substituted into a formula of a preset score card model to calculate values of constants A and B, wherein the P and the PDO can be any value between positive infinity and negative infinity. And calculating credit scores corresponding to the credit probabilities of the target users according to the preset score card model, and judging whether the target users can loan, loan amount or repayment strategy according to the credit scores.
206. And counting credit scores and whether the system users violate the constraint information, and calculating a violation threshold score.
Whether the information is violated refers to information that the user of the system has already violated or not violated with the completed financial transaction record. The credit score of the system user and whether the information is violated are counted, and a threshold value score of the violation is calculated, namely, when the credit score of the system user exceeds a certain score, the system user does not violate the violation.
207. And if the credit score is larger than the default threshold score, calculating the loan amount of the target user according to the credit score.
And when the system user does not have the default risk or the default risk is larger, calculating the loan amount of the target user according to the credit score. Before calculating the loan amount, the highest loan amount corresponding to the maximum information score can be calculated, and the loan amount of the target user is calculated in proportion according to the maximum credit score and the credit score of the target user.
The invention provides a credit evaluation method, which comprises the steps of firstly obtaining a user track label of each system user, then constructing a knowledge graph according to a user interest label, a user track label and a user call record of each system user, extracting a user feature vector according to the knowledge graph and a Node2vec algorithm, inputting user information of a target user into a preset neural network model, calculating the credit probability of whether the target user violates, and finally inputting the credit probability into a preset scoring card model to calculate the credit score of the target user. Compared with the prior art, the method and the device have the advantages that through the knowledge graph constructed by the user track labels, the user interest labels and the user call records, the association relation between the users is extracted for credit evaluation, the credit of the individuals is evaluated in the partition group, the accuracy of evaluation data for credit evaluation is improved, and the effectiveness of the credit score obtained through calculation is improved. And extracting the user feature vector in the knowledge graph by using a Node2vec algorithm, so that the feature accuracy of the user feature vector reaction can be increased. The credit probability of the user is predicted through the neural network model, so that the credit score is more accurate, and the personal financial rights of the user such as the credit line, the credit card passing rate, the application amount and the like can be accurately estimated through the credit score.
Further, as an implementation of the method shown in fig. 1, an embodiment of the present invention provides a credit evaluation apparatus, as shown in fig. 3, where the apparatus includes:
an obtaining module 31, configured to obtain a user track label according to location-based service LBS data of each system user;
a construction module 32, configured to construct a knowledge graph according to the user interest tag, the user track tag and the user call record of each system user, where the knowledge graph is used to identify the association relationship between all the system users;
the extracting module 33 is configured to extract a user feature vector according to the knowledge graph by using a Node2vec algorithm;
the calculating module 34 is configured to input user information of a target user into a preset neural network model, and calculate a credit probability of whether the target user violates, where the user information includes user personal information of the target user, a user interest tag of the target user, a user track tag of the target user, the user feature vector, and a user call record of the target user;
the calculation module 34 is further configured to input the credit probability into a preset scoring card model, and calculate a credit score of the target user.
The invention provides a credit evaluation device, which comprises the steps of firstly acquiring a user track label of each system user, then constructing a knowledge graph according to a user interest label, a user track label and a user call record of each system user, extracting a user feature vector according to the knowledge graph and a Node2vec algorithm, inputting user information of a target user into a preset neural network model, calculating whether the target user violates credit probability, and finally inputting the credit probability into a preset scoring card model to calculate the credit score of the target user. Compared with the prior art, the method and the device have the advantages that through the knowledge graph constructed by the user track labels, the user interest labels and the user call records, the association relation between the users is extracted for credit evaluation, the credit of the individuals is evaluated in the partition group, the accuracy of evaluation data for credit evaluation is improved, and the effectiveness of the credit score obtained through calculation is improved. And extracting the user feature vector in the knowledge graph by using a Node2vec algorithm, so that the feature accuracy of the user feature vector reaction can be increased. The credit probability of the user is predicted through the neural network model, so that the credit score is more accurate, and the personal financial rights of the user such as the credit line, the credit card passing rate, the application amount and the like can be accurately estimated through the credit score.
Further, as an implementation of the method shown in fig. 2, another apparatus for credit assessment is provided in an embodiment of the present invention, as shown in fig. 4, where the apparatus includes:
an obtaining module 41, configured to obtain a user track label according to location-based service LBS data of each system user;
a construction module 42, configured to construct a knowledge graph according to the user interest tag, the user track tag and the user call record of each system user, where the knowledge graph is used to identify the association relationship between all the system users;
an extracting module 43, configured to extract a user feature vector according to the knowledge graph by using a Node2vec algorithm;
a calculation module 44, configured to input user information of a target user into a preset neural network model, and calculate a credit probability of whether the target user violates, where the user information includes user personal information of the target user, a user interest tag of the target user, a user track tag of the target user, the user feature vector, and a user call record of the target user;
the calculating module 44 is further configured to input the credit probability into a preset scoring card model, and calculate a credit score of the target user.
Further, the obtaining module 41 includes:
an acquisition unit 411, configured to sequentially acquire LBS data in each system user according to a preset frequency, where the LBS data is represented by longitude and latitude coordinates;
an obtaining unit 412, configured to obtain, in the electronic map, a map marking location within a preset distance range with the LBS data as a coordinate center point;
and the determining unit 413 is configured to determine the map marking point as a user track label corresponding to the system user.
Further, the building block 42 comprises:
a searching unit 421, configured to search for the same associated user as the user interest tag, the user track tag, and the user call record of each of the system users;
a building unit 422, configured to build a mapping relationship between each system user and the associated user;
and a construction unit 423, configured to construct the knowledge graph according to the mapping relationship.
Further, the extracting module 43 includes:
a calculating unit 431, configured to sequentially use each system user as a target user, and calculate, using a Node2vec algorithm, a default probability of the target user;
and the generating unit 432 is configured to aggregate the default probabilities, and generate the user feature vector in a combined manner.
Further, the calculating unit 431 includes:
an obtaining subunit 4311, configured to obtain, in the knowledge graph, a preset number of node users corresponding to the target user according to a random walk algorithm, where the node users and the target user have an association relationship, and the association relationship includes a first-level association relationship directly connected between the target user and the node user, and a first-level second-level association relationship indirectly connected between the target user and the node user;
a searching subunit 4332, configured to search whether the loan user in the node user violates the constraint information;
and a calculating subunit 4333, configured to calculate the default probability of the target user according to the default information and the association weight corresponding to the association relationship.
Further, the user information further comprises a label frequency and a label correlation coefficient of a user track label of the target user, and/or a social track label of the target user, a label frequency and a label correlation coefficient of the social track label;
the method further comprises the steps of:
the statistics module 45 is configured to, before inputting the user information of the target user into a preset neural network model and calculating the confidence probability of whether the target user violates, count the label frequency of the user track label of the target user in a preset time range, and calculate the label correlation coefficient of the user track label according to a Pearson correlation coefficient formula, where the Pearson correlation coefficient formula is Wherein ρ is X,Y Is the label correlation coefficient of the user track label, X is the user track label, Y is the system user default information with the user track label, cov (X, Y) is the covariance between X and Y, sigma X Is the standard deviation of X, sigma Y Is the standard deviation of Y; and/or the number of the groups of groups,
the statistics module 45 is further configured to obtain a social track label corresponding to social LBS data reported by a social media APP according to authorization information of the target user, count a label frequency of the social track label, and calculate a label correlation coefficient of the social track label according to the Pearson correlation coefficient formula, where the Pearson correlation coefficient formula isWherein ρ is X,Y Is the tag correlation coefficient of the social track tag, X is the social track tag, Y is the information whether the system user with the social track tag violates the constraint, cov (X, Y) is the covariance between X and Y, sigma X Is the standard deviation of X, sigma Y Is the standard deviation of Y.
Further, the method further comprises:
the calculation module 44 is further configured to input the credit probability into a preset scoring card model, calculate a credit score of the target user, calculate a credit score of the system user and whether to violate the constraint information, and calculate a default threshold score;
The calculating module 44 is further configured to calculate a loan amount of the target user according to the credit score if the credit score is greater than the default threshold score.
The invention provides a credit evaluation device, which comprises the steps of firstly acquiring a user track label of each system user, then constructing a knowledge graph according to a user interest label, a user track label and a user call record of each system user, extracting a user feature vector according to the knowledge graph and a Node2vec algorithm, inputting user information of a target user into a preset neural network model, calculating whether the target user violates credit probability, and finally inputting the credit probability into a preset scoring card model to calculate the credit score of the target user. Compared with the prior art, the method and the device have the advantages that through the knowledge graph constructed by the user track labels, the user interest labels and the user call records, the association relation between the users is extracted for credit evaluation, the credit of the individuals is evaluated in the partition group, the accuracy of evaluation data for credit evaluation is improved, and the effectiveness of the credit score obtained through calculation is improved. And extracting the user feature vector in the knowledge graph by using a Node2vec algorithm, so that the feature accuracy of the user feature vector reaction can be increased. The credit probability of the user is predicted through the neural network model, so that the credit score is more accurate, and the personal financial rights of the user such as the credit line, the credit card passing rate, the application amount and the like can be accurately estimated through the credit score.
According to one embodiment of the present invention, there is provided a storage medium storing at least one executable instruction for performing the method of credit assessment in any of the method embodiments described above.
Fig. 5 is a schematic structural diagram of a computer device according to an embodiment of the present invention, and the specific embodiment of the present invention is not limited to the specific implementation of the computer device.
As shown in fig. 5, the computer device may include: a processor 502, a communication interface (Communications Interface) 504, a memory 506, and a communication bus 508.
Wherein: processor 502, communication interface 504, and memory 506 communicate with each other via communication bus 508.
A communication interface 504 for communicating with network elements of other devices, such as clients or other servers.
The processor 502 is configured to execute the program 510, and may specifically perform relevant steps in the above-described method embodiment of credit evaluation.
In particular, program 510 may include program code including computer-operating instructions.
The processor 502 may be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement embodiments of the present invention. The one or more processors included in the computer device may be the same type of processor, such as one or more CPUs; but may also be different types of processors such as one or more CPUs and one or more ASICs.
A memory 506 for storing a program 510. Memory 506 may comprise high-speed RAM memory or may also include non-volatile memory (non-volatile memory), such as at least one disk memory.
The program 510 may be specifically operable to cause the processor 502 to:
acquiring user track labels according to location-based service (LBS) data of each system user;
constructing a knowledge graph according to the user interest labels, the user track labels and the user call records of each system user, wherein the knowledge graph is used for identifying the association relation among all the system users;
extracting user feature vectors by using a Node2vec algorithm according to the knowledge graph;
inputting user information of a target user into a preset neural network model, and calculating credit probability of whether the target user breaks down, wherein the user information comprises user personal information of the target user, user interest labels of the target user, user track labels of the target user, user feature vectors and user call records of the target user;
and inputting the credit probability into a preset scoring card model, and calculating the credit score of the target user.
It will be appreciated by those skilled in the art that the modules or steps of the invention described above may be implemented in a general purpose computing device, they may be concentrated on a single computing device, or distributed across a network of computing devices, they may alternatively be implemented in program code executable by computing devices, so that they may be stored in a memory device for execution by computing devices, and in some cases, the steps shown or described may be performed in a different order than that shown or described, or they may be separately fabricated into individual integrated circuit modules, or multiple modules or steps within them may be fabricated into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (9)
1. A method of credit assessment, comprising:
Acquiring user track labels according to location-based service (LBS) data of each system user;
constructing a knowledge graph according to the user interest labels, the user track labels and the user call records of each system user, wherein the knowledge graph is used for identifying the association relation among all the system users;
extracting user feature vectors by using a Node2vec algorithm according to the knowledge graph;
inputting user information of a target user into a preset neural network model, and calculating credit probability of whether the target user breaks down, wherein the user information comprises user personal information of the target user, user interest labels of the target user, user track labels of the target user, user feature vectors and user call records of the target user;
inputting the credit probability into a preset scoring card model, and calculating the credit score of the target user;
and extracting the user characteristic vector by using a Node2vec algorithm according to the knowledge graph, wherein the method comprises the following steps:
sequentially taking each system user as a target user, and calculating the default probability of the target user based on the default information of each Node user corresponding to the target user and the association weight of the association relation between the target user and each Node user by using a Node2vec algorithm;
And collecting the default probabilities, and combining to generate the user feature vector, wherein the user feature vector comprises the default probabilities of all system users.
2. The method of claim 1, wherein the obtaining the user trajectory tag according to the location based service LBS data of each system user comprises:
according to preset frequency, sequentially collecting LBS data in each system user, wherein the LBS data are represented by longitude and latitude coordinates;
in the electronic map, taking the LBS data as a coordinate center point to obtain map marking places within a preset distance range;
and determining the map marking points as user track labels corresponding to the system users.
3. The method of claim 1, wherein constructing a knowledge graph from the user interest tags, the user trajectory tags, and the user call records for each of the system users comprises:
searching the associated users which are the same as the user interest labels, the user track labels and the user call records of each system user;
establishing a mapping relation between each system user and the associated user;
and constructing the knowledge graph according to the mapping relation.
4. The method of claim 1, wherein the calculating, using the Node2vec algorithm, the breach probability of the target user based on breach information of each Node user corresponding to the target user and association weights of the association relationships with each Node user for the target, comprises:
in the knowledge graph, according to a random walk algorithm, obtaining a preset number of node users corresponding to the target user, wherein the node users and the target user have an association relationship, and the association relationship comprises a primary association relationship directly connected between the target user and the node user and a secondary association relationship indirectly connected between the target user and the node user;
searching whether the loan user in the node users violates the constraint information;
and calculating the default probability of the target user according to the default information and the association weight corresponding to the association relation.
5. The method of claim 1, wherein the user information further comprises a tag frequency and a tag correlation coefficient of a user track tag of the target user, and/or a social track tag of the target user, a tag frequency and a tag correlation coefficient of the social track tag;
Before the user information of the target user is input into a preset neural network model and the credit probability of whether the target user violates is calculated, the method further comprises:
in a preset time range, counting the label frequency of the user track label of the target user, and calculating the label correlation coefficient of the user track label according to a Pearson correlation coefficient formula, wherein the Pearson correlation coefficient formula is thatWherein ρ is X,Y Is the label correlation coefficient of the user track label, X is the user track label, Y is the system user default information with the user track label, cov (X, Y) is the covariance between X and Y, sigma X Is the standard deviation of X, sigma Y Is the standard deviation of Y; and/or the number of the groups of groups,
acquiring a social track label corresponding to social LBS data reported by a social media APP according to authorization information of the target user, counting the label frequency of the social track label, and calculating a label correlation coefficient of the social track label according to a Pearson correlation coefficient formula, wherein the Pearson correlation coefficient formula is as followsWherein ρ is X,Y Is the tag correlation coefficient of the social track tag, X is the social track tag, Y is the information whether the system user with the social track tag violates the constraint, cov (X, Y) is the covariance between X and Y, sigma X Is the standard deviation of X, sigma Y Is the standard deviation of Y.
6. The method of claim 1, wherein said inputting said credit probability into a pre-set scoring card model, after calculating a credit score for said target user, said method further comprises:
counting credit scores and whether the system users violate the constraint information, and calculating a violation threshold score;
and if the credit score is larger than the default threshold score, calculating the loan amount of the target user according to the credit score.
7. An apparatus for credit assessment, comprising:
the acquisition module is used for acquiring user track labels according to the LBS data of each system user based on the location service;
the building module is used for building a knowledge graph according to the user interest labels, the user track labels and the user call records of each system user, and the knowledge graph is used for identifying the association relation among all the system users;
the extraction module is used for extracting user feature vectors by using a Node2vec algorithm according to the knowledge graph;
the calculation module is used for inputting user information of a target user into a preset neural network model, and calculating credit probability of whether the target user breaks down, wherein the user information comprises user personal information of the target user, user interest labels of the target user, user track labels of the target user, user feature vectors and user call records of the target user;
The calculation module is also used for inputting the credit probability into a preset scoring card model and calculating the credit score of the target user;
the extraction module comprises a calculation unit and a generation unit:
the computing unit is used for sequentially taking each system user as a target user, and computing the default probability of the target user based on the default information of each Node user corresponding to the target user and the association weight of the target for the association relation with each Node user by using a Node2vec algorithm;
the generating unit is used for gathering the default probabilities and generating the user feature vector in a combined mode, and the user feature vector comprises the default probabilities of all system users.
8. A storage medium having stored therein at least one executable instruction for causing a processor to perform operations corresponding to the method of credit assessment of any of claims 1-6.
9. A computer device, comprising: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;
the memory is configured to store at least one executable instruction that causes the processor to perform operations corresponding to the method of credit assessment of any of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910780782.8A CN110689423B (en) | 2019-08-22 | Credit evaluation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910780782.8A CN110689423B (en) | 2019-08-22 | Credit evaluation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110689423A CN110689423A (en) | 2020-01-14 |
| CN110689423B true CN110689423B (en) | 2023-08-11 |
Family
ID=
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107943879A (en) * | 2017-11-14 | 2018-04-20 | 上海维信荟智金融科技有限公司 | Fraud group detection method and system based on social networks |
| CN109242673A (en) * | 2018-11-04 | 2019-01-18 | 上海良鑫网络科技有限公司 | Hawkeye is counter to cheat big data air control assessment system |
| CN109657837A (en) * | 2018-11-19 | 2019-04-19 | 平安科技（深圳）有限公司 | Default Probability prediction technique, device, computer equipment and storage medium |
| CN109685647A (en) * | 2018-12-27 | 2019-04-26 | 阳光财产保险股份有限公司 | The training method of credit fraud detection method and its model, device and server |
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107943879A (en) * | 2017-11-14 | 2018-04-20 | 上海维信荟智金融科技有限公司 | Fraud group detection method and system based on social networks |
| CN109242673A (en) * | 2018-11-04 | 2019-01-18 | 上海良鑫网络科技有限公司 | Hawkeye is counter to cheat big data air control assessment system |
| CN109657837A (en) * | 2018-11-19 | 2019-04-19 | 平安科技（深圳）有限公司 | Default Probability prediction technique, device, computer equipment and storage medium |
| CN109685647A (en) * | 2018-12-27 | 2019-04-26 | 阳光财产保险股份有限公司 | The training method of credit fraud detection method and its model, device and server |
Non-Patent Citations (1)
| Title |
|---|
| 一种知识图谱的排序学习个性化推荐算法;杨晋吉;胡波;王欣明;伍昱燊;赵淦森;;小型微型计算机***(第11期);第69-73页 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210329094A1 (en) | Discovering signature of electronic social networks | |
| CN110009174B (en) | Risk recognition model training method and device and server | |
| CN105512938A (en) | Online credit risk assessment method based on long-term using behavior of user | |
| CN107146089A (en) | The single recognition methods of one kind brush and device, electronic equipment | |
| CN108833453A (en) | A kind of method and apparatus determined using account | |
| CN103761254A (en) | Method for matching and recommending service themes in various fields | |
| CN108153824A (en) | The determining method and device of targeted user population | |
| CN107153656A (en) | A kind of information search method and device | |
| CN107657357B (en) | Data processing method and device | |
| CN112632409A (en) | Same user identification method, device, computer equipment and storage medium | |
| CN107622406A (en) | Identify the method and system of virtual unit | |
| CN112990989B (en) | Value prediction model input data generation method, device, equipment and medium | |
| CN114331698A (en) | Risk portrait generation method and device, terminal and storage medium | |
| CN112347457A (en) | Abnormal account detection method and device, computer equipment and storage medium | |
| CN113887214A (en) | Artificial intelligence based wish presumption method and related equipment thereof | |
| CN113850669A (en) | User grouping method and device, computer equipment and computer readable storage medium | |
| CN113420789A (en) | Method, device, storage medium and computer equipment for predicting risk account | |
| CN111476371A (en) | Method and device for evaluating specific risk faced by server | |
| CN110689423B (en) | Credit evaluation method and device | |
| CN111222923A (en) | Method and device for judging potential customer, electronic equipment and storage medium | |
| CN109636378B (en) | Account identification method and device and electronic equipment | |
| JPWO2021048902A1 (en) | Learning model application system, learning model application method, and program | |
| CN110570301B (en) | Risk identification method, device, equipment and medium | |
| CN111476511B (en) | Data display method and device for specific risks faced by server | |
| CN110689423A (en) | Credit assessment method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |