WO2020233360A1

WO2020233360A1 - Method and device for generating product evaluation model

Info

Publication number: WO2020233360A1
Application number: PCT/CN2020/087382
Authority: WO
Inventors: 胡甜敏
Original assignee: 深圳壹账通智能科技有限公司
Priority date: 2019-05-22
Filing date: 2020-04-28
Publication date: 2020-11-26
Also published as: CN110264038A

Abstract

Provided is a method for generating a product evaluation model. The method comprises: on the basis of preset derived variable conversion functions respectively corresponding to different primitive variables, configuring a plurality of derived variables for each of the primitive variables in a variable base, and establishing correlations between the primitive variables and the derived variables; acquiring product information of a target product, and performing semantic analysis on the product information to identify a product keyword included in the product information; selecting, from the variable base, primitive variables matching the product keyword, and identifying the selected primitive variables as target variables; on the basis of the correlation, acquiring a derived variable associated with each of the target variables; and according to the product type of the target product, downloading a product evaluation template of the product type, and importing the target variables and the derived variables associated with the target variables into the product evaluation template to generate a product evaluation model of the target product. According to the present application, various variables associated with the target product can be directly extracted from the variable base, thereby improving the efficiency of generating an evaluation model.

Description

Method and equipment for generating product evaluation model

This application claims the priority of a Chinese patent application filed with the Chinese Patent Office on May 22, 2019. The application number is 201910430235.7 and the invention title is "A method and equipment for generating a product evaluation model". The entire content is incorporated by reference. In this application.

Technical field

This application belongs to the technical field of model management, and in particular relates to a method and equipment for generating a product evaluation model.

Background technique

Before the user conducts risk operations, the company needs to conduct risk evaluation on the user and determine the user evaluation level of the user. The evaluation models of different products are different. As the number of products continues to increase, the frequency of product evaluation models is also increased. How to build a product evaluation model efficiently will directly affect the update speed of the product. Existing product evaluation models include not only the native variables directly extracted from user information, but also the derivative variables obtained through the established conversion algorithm based on the native variables. Users need to manually configure the native variables and derivative variables included in each risk model. The number of variables that need to be configured is huge, and variables often miss the configuration, which reduces the efficiency of creating the product evaluation model and the accuracy of the model.

technical problem

In view of this, the embodiments of the present application provide a method and equipment for generating product evaluation models to solve the existing technology for generating product evaluation models. It is necessary to manually configure the native variables and derivative variables contained in each risk model. When the number is large, the variables are often omitted from configuration, which reduces the efficiency of creating the product evaluation model and the accuracy of the model.

Technical solutions

The first aspect of the embodiments of the present application provides a method for generating a product evaluation model, including: configuring multiple derivative variables for each of the native variables in the variable library based on preset derivative variable conversion functions corresponding to different native variables. , And establish the corresponding relationship between the native variable and the derived variable; obtain product information of the target product, perform semantic analysis on the product information, and identify the product keywords contained in the product information; from the variable library Select the native variable that matches the product keyword, and identify the selected native variable as the target variable corresponding to the target product; based on the corresponding relationship, obtain the derivative associated with each target variable Variable; according to the product type of the target product, download the product evaluation template of the product type, and import the target variable and the derivative variable associated with the target variable into the product evaluation template to generate the target Product evaluation model of the product.

The second aspect of the embodiments of the present application provides a device for generating a product evaluation model, including: a derivative variable configuration unit, which is configured to correspond to preset derivative variable conversion functions corresponding to different native variables, which are each described in the variable library. The native variable configures multiple derivative variables, and establishes the corresponding relationship between the native variable and the derivative variable; the product keyword acquisition unit is used to obtain product information of the target product, and perform semantic analysis on the product information to identify all The product keywords included in the product information; a target variable identification unit for selecting the native variable matching the product keyword from the variable library, and identifying the selected native variable as the target product Corresponding target variable; a derivative variable selection unit for obtaining the derivative variable associated with each of the target variables based on the corresponding relationship; a product evaluation model generating unit for downloading the target product according to the product type of the target product Describe the product evaluation template of the product type, and import the target variable and the derivative variable associated with the target variable into the product evaluation template to generate a product evaluation model of the target product.

The third aspect of the embodiments of the present application provides a terminal device, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, Realize the steps of the first aspect.

A fourth aspect of the embodiments of the present application provides a computer-readable storage medium that stores a computer program that implements the steps of the first aspect when the computer program is executed by a processor.

Beneficial effect

Implementing the method and device for generating a product evaluation model provided by the embodiments of this application has the following beneficial effects:

The embodiment of the application first configures the associated derivative variables for each native variable, and establishes the corresponding relationship between each native variable and the derivative variable; then analyzes the product information of the target product to determine the native variable contained in the target product, based on the above The established correspondence relationship automatically pulls the derivative variables associated with each native variable, and imports the obtained native variables and derivative variables into the product evaluation template that matches the product type of the target product to generate the product evaluation model of the target product, and realize automatic generation The purpose of the product evaluation model. Compared with the existing product evaluation model generation technology, users do not need to manually configure native variables, and configure a conversion function for each native variable, so as to obtain derivative variables. The variables associated with the target product can be directly extracted from the variable library to improve The generation efficiency of the evaluation model can also avoid the occurrence of missing variables and improve the accuracy of the product evaluation model.

Description of the drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only of the present application. For some embodiments, for those of ordinary skill in the art, other drawings can be obtained from these drawings without creative labor.

FIG. 1 is an implementation flowchart of a method for generating a product evaluation model provided by the first embodiment of the present application;

2 is a specific implementation flowchart of a method S101 for generating a product evaluation model provided by the second embodiment of the present application;

3 is a specific implementation flowchart of a method for generating a product evaluation model provided by the third embodiment of the present application;

4 is a specific implementation flowchart of a method S103 for generating a product evaluation model provided by the fourth embodiment of the present application;

FIG. 5 is a specific implementation flowchart of a method for generating a product evaluation model provided by the fifth embodiment of the present application;

6 is a structural block diagram of a device for generating a product evaluation model provided by an embodiment of the present application;

Fig. 7 is a schematic diagram of a terminal device provided by another embodiment of the present application.

Embodiments of the invention

In order to make the purpose, technical solutions, and advantages of this application clearer, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the application, and are not used to limit the application.

The embodiment of the application first configures the associated derivative variables for each native variable, and establishes the corresponding relationship between each native variable and the derivative variable; then analyzes the product information of the target product to determine the native variable contained in the target product, based on the above The established correspondence relationship automatically pulls the derivative variables associated with each native variable, and imports the obtained native variables and derivative variables into the product evaluation template that matches the product type of the target product to generate the product evaluation model of the target product, and realize automatic generation The purpose of the product evaluation model is to solve the existing technology of product evaluation model generation. It is necessary to manually configure the native variables and derived variables contained in each risk model. When the number of native variables is large, the variables are often omitted from configuration, thereby reducing The efficiency of the creation of the product evaluation model and the accuracy of the model are addressed.

In the embodiment of the present application, the execution subject of the process is the terminal device. The terminal equipment includes, but is not limited to: servers, computers, smart phones, tablet computers, and other equipment capable of performing the operation of generating product evaluation models. Fig. 1 shows an implementation flow chart of the method for generating a product evaluation model provided by the first embodiment of the present application, which is detailed as follows:

In S101, based on preset derivative variable conversion functions corresponding to different native variables, a plurality of derivative variables are configured for each native variable in the variable library, and the corresponding relationship between the native variable and the derivative variable is established.

In this embodiment, the terminal device needs to establish a variable library before creating a product evaluation model for each target product. The variable library contains native variables and derived variables generated through preset conversion functions based on each native variable. Different native variables correspond to different derivative variable conversion functions, and the terminal device can extract all derivative variable conversion functions related to the native variable to establish the correspondence between the native variable and its corresponding derivative variable. Among them, the native variables are specifically parameters that can be directly extracted from the original data. The original data may be user information for purchasing the target product and/or product information of the target product. Derivative variables can be calculated by a single native variable based on a preset calculation method. For example, a native variable is the "annual income" of the user under evaluation who purchases the target product, that is, the "annual income" of the user under evaluation can be calculated based on the "annual income". "Monthly income", because the information of monthly income cannot be directly extracted from user information, it needs to be obtained after calculating annual income. Therefore, "monthly income" is a derivative variable of the original variable of "annual income". Of course, the derivative variable can also be calculated from two or more native variables through a preset conversion function. For example, one native variable is "deposit amount" and the other native variable is "shopping business amount". "Deposit amount" and "Shopping business amount" determine the "total assets" of users who purchase the target product. Similarly, "total assets" cannot be directly extracted from user information. You must first determine the "deposit amount" and "shopping Business amount" is calculated based on the above two parameters, that is, "total assets" is a derivative variable of "deposit amount" and "shopping business amount". It can be seen that native variables can correspond to multiple derivative variables. When there are a large number of native variables, if you pull the derivative variables manually by manual configuration, you may miss the derivative variables based on the conversion of multiple native variables. It is necessary to consider the derivative variables corresponding to a single native variable, and it is also necessary to determine whether there are derivative variables in different combinations of native variables, which increases the difficulty of configuration and causes variables to be omitted.

In this embodiment, the terminal device can receive the derivative variable conversion function input by the user, and identify the native variables and derivative variables included in the derivative variable conversion function, and recognize the aforementioned native variables as the native variables to be configured, and establish the The corresponding relationship between the configured native variable and the derivative variable, preferably, the above-mentioned derivative variable conversion function is imported into the corresponding relationship, so that the conversion function of the derivative variable can be directly determined from the corresponding relationship, and the created value is associated with the native variable Import the derived variables into the variable library.

Optionally, in this embodiment, the terminal device may send a file containing the derivative variable conversion function to the terminal device, and the terminal device parses the file, separates multiple derivative variable conversion functions, and identifies the derivative variable function In the independent variable and dependent variable, the variable name corresponding to the independent variable is identified as the original variable, and the variable name of the dependent variable is identified as the derived variable, so as to determine the corresponding relationship between the original variable and the derived variable.

In S102, product information of the target product is obtained, and semantic analysis is performed on the product information to identify product keywords contained in the product information.

In this embodiment, after the terminal device has established the variable library, it can create a product evaluation model for the target product. The product evaluation model is specifically used to determine the user level required to purchase the target product according to the product characteristics of the target product, and determine whether to respond to the user's purchase operation based on the user level, that is, to determine whether the user has the purchase authority of the target product For example, the product evaluation model is specifically a risk evaluation model. The risk evaluation model can be used to determine the purchase user’s risk level, and based on the risk level determine the purchase user’s loanable amount range, and determine whether the purchase user’s loan product purchase Within the range of the loanable amount, if it is within the range of the acceptable amount, the user will respond to this purchase operation; on the contrary, if the purchase user’s loan amount is outside the range of the acceptable amount, the purchase operation will be identified as a failure . Based on this, the terminal device needs to determine the user information content and/or product information content of the target product that needs to be collected, so as to be able to determine the legality of the user or operation through the collected parameters.

In this embodiment, the terminal device extracts the product information associated with the product identifier from the product database according to the product identifier of the target product. The product information may be a description document or a development document of the target product. Preferably, the product information may be a collection page of user information, and the collection page of user information contains the names of various parameters required to collect user information. After obtaining the product information of the target product, the terminal device performs semantic analysis on the product information, determines the degree of association of each character in the product information, selects multiple characters with a higher degree of association to form candidate keywords, and based on the candidate key The word attribute of the word, and the product keywords are extracted from the candidate keywords. Optionally, the terminal device can store different native variables configured with related product keywords. The terminal device aggregates the product keywords corresponding to all native variables to generate a product keyword library. When performing the product keyword extraction operation, It will detect whether the product information contains the product keywords recorded in the product keyword database. If the product information includes the keywords recorded in the product keyword database, then the keyword will be identified as the product keyword of the target product.

In S103, the native variable matching the product keyword is selected from the variable library, and the selected native variable is identified as the target variable corresponding to the target product.

In this embodiment, after the terminal device determines the product keywords contained in the product information, it can determine the type of parameters that need to be collected when the user who purchases the target product performs the evaluation operation, so the product keywords need to be converted into evaluation The variable form that the model can read. The terminal device matches the product keyword with each native variable in the variable library, and selects the native variable whose matching degree is higher than the preset matching threshold as the native variable associated with the product keyword. Optionally, if the matching degree between the product keyword and two or more native variables is greater than a preset variable threshold, the native variable with the largest matching degree is selected as the native variable associated with the candidate keyword.

In this embodiment, the method of identifying the native variable matching the product keyword may be: the terminal device obtains the variable name of each native variable, matches each variable name with the product keyword, and selects the native variable whose name is the same as the product keyword. Variables are used as native variables for product keyword matching.

In this embodiment, the terminal device uses the native variable matching the product keyword as the target variable corresponding to the target product, and generates a product evaluation model of the target product through the target variable.

In S104, the derivative variable associated with each target variable is obtained based on the corresponding relationship.

In this embodiment, since the terminal device has determined multiple derived variables of each native variable, the terminal device can determine the corresponding relationship of the target variable according to the variable name of the target variable, and extract the corresponding relationship from the variable library through the corresponding relationship. The derived variables associated with the target variable and the conversion function of each derived variable.

In S105, according to the product type of the target product, download the product evaluation template of the product type, and import the target variable and the derivative variable associated with the target variable into the product evaluation template to generate all Describe the product evaluation model of the target product.

In this embodiment, the terminal device needs to obtain the product evaluation template of the target product after determining the target variable required for the evaluation of the target product and the derivative variable associated with the target variable. Due to different product types, the evaluation calculation method will be different. For example, for financial products, the user’s credit rating needs to be determined, and the product evaluation template corresponding to the product type is the credit evaluation template; for insurance products, the user’s risk needs to be determined Level, the product evaluation template corresponding to the product type is the risk evaluation template. Therefore, in order to match the generated product evaluation model with the target product, the terminal device needs to determine the product type of the target product, so as to select the product evaluation template associated with the product type. Optionally, the method for the terminal device to identify the product type of the target product may be: the terminal device obtains the module list of the target product, and determines the product type of the target product based on the module function of each program module in the module list. Each module function corresponds to the probability of a product type. The terminal device weights each module function and selects the product type with the highest probability as the product type of the target product.

In this embodiment, after determining the product type of the target product, the terminal device can obtain a product evaluation template that matches the product type, and collect the variables required by the target product, namely the target variable and the derivative associated with the target variable. The variables are also imported into the product evaluation template, and the user level of the user who purchases the target product can be determined based on the above variable information.

It can be seen from the above that the method for generating a product evaluation model provided by the embodiments of this application first configures the associated derivative variables for each native variable, and establishes the corresponding relationship between each native variable and the derivative variable; then, the product information of the target product Perform analysis to determine the native variables contained in the target product, and based on the corresponding relationship established above, automatically pull the derived variables associated with each native variable, and import the obtained native variables and derived variables into products that match the product type of the target product In the evaluation template, a product evaluation model of the target product is generated to realize the purpose of automatically generating a product evaluation model. Compared with the existing product evaluation model generation technology, users do not need to manually configure native variables, and configure a conversion function for each native variable, so as to obtain derivative variables. The variables associated with the target product can be directly extracted from the variable library to improve The generation efficiency of the evaluation model can also avoid the occurrence of missing variables and improve the accuracy of the product evaluation model

FIG. 2 shows a specific implementation flowchart of a method S101 for generating a product evaluation model provided by the second embodiment of the present application. Referring to FIG. 2, compared to the embodiment described in FIG. 1, the method S101 for generating a product evaluation model provided by this embodiment includes: S1011 to S1014, which are detailed as follows:

Further, based on the preset derivative variable conversion functions corresponding to different native variables, configure multiple derivative variables for each native variable in the variable library, and establish the corresponding relationship between the native variable and the derivative variable, include:

In S1011, the created in-use evaluation model is analyzed, and the native variables and the derived variables included in the in-use evaluation model are determined.

In this embodiment, in addition to obtaining the derivative variable conversion function through manual configuration by the user, the terminal device may also obtain the conversion function between the native variable and the derivative variable in the historically generated product evaluation model through intelligent learning. Based on this, the terminal device can extract the created in-use evaluation model from the evaluation model library, and the in-use evaluation model is specifically the product evaluation model corresponding to the historical product. The in-use evaluation model can automatically pull the parameter value of the native variable from the user information, output the parameter value of the derivative variable based on the parameter value of each native variable, and calculate the product measurement score of the historical product based on the native variable and the derivative parameter. The terminal device can determine the conversion weight of each native variable to the derivative variable through the control variable method, thereby determining the conversion function between the two.

In this embodiment, the terminal device can analyze the in-use evaluation model, identify the list of variables included in the in-use evaluation model, and divide the original variables according to the conversion relationship between each variable and user information and/or product information And derived variables.

In 1012, the parameter value range of the native variable included in the in-use evaluation model is obtained, a plurality of parameter nodes are selected within the parameter value range, and a training variable value is configured for each parameter node.

In this embodiment, after determining the native variable in the in-use evaluation model, the terminal device determines the data type of the native variable according to the variable type, and determines the parameter value range based on the data type. For example, if a native variable is "gender" information, the data type is character type, and the value range is "male", "female" and "third gender", each value is identified as a parameter node, and Configure a training variable value for each parameter node; for another example, if a native variable is "age" information, the data type is numeric, and the value range is [0,120], the terminal device can select more from the age information. Key age nodes, such as "18", "25", "50", "70", and configure the corresponding training variable value for each key age node.

Among them, for a discrete limited value range, each discrete value can be used as a parameter node, and for a continuous infinite value range, you can select multiple eigenvalues according to the variable attributes of the original variable, and use each eigenvalue as A parameter node.

In S1013, each of the training variable values is imported into the in-use evaluation model, and the first derivative variable values output by the in-use evaluation model based on the training variable values are collected.

In this embodiment, the terminal device can use the control variable method, that is, when importing the training variable values of a native variable into the in-use evaluation model, keeping other native variables except the native variable at a fixed value, so as to determine the current When the original variable changes, the changed derivative variable can determine the derivative variable associated with the original variable, and obtain the first derivative variable value output by the associated derivative variable based on the value of the training variable.

In S1014, according to the value of the training variable and the value of the first derivative variable, the derivative variable conversion function corresponding to the native variable and the derivative variable included in the in-use evaluation model is determined.

In this embodiment, the terminal device can determine the conversion weight of the native variable to the derivative variable according to the multiple training variable values and the first derivative variable value with the training variable value, thereby generating the conversion between the native variable and the derivative variable. function. If the derivative variable needs to be calculated through the conversion of multiple native variables, the conversion weight between the derivative variable and the native variable is obtained after weighting based on the conversion weight corresponding to the multiple native variable, that is, the aforementioned derivative variable conversion function.

In the embodiment of the present application, the conversion function between the native variable and the derived variable is determined by analyzing the evaluation model in use, thereby realizing the purpose of automatically learning the conversion function without manual configuration by the user, which improves the generation of the product evaluation model effectiveness.

Fig. 3 shows a specific implementation flow chart of a method for generating a product evaluation model provided by the third embodiment of the present application. Referring to FIG. 3, compared to the embodiment described in FIG. 1, the method for generating a product evaluation model provided by this embodiment includes importing the target variable and the derivative variable associated with the target variable into the The product evaluation template, after generating the product evaluation model of the target product, also includes: S301~S304, the details are as follows:

In S301, historical information of a plurality of evaluated users is obtained; the historical information includes historical user parameters and historical evaluation levels of the evaluated users.

In this embodiment, after the terminal device generates the product evaluation model of the target product, it needs to check the product evaluation model, that is, to determine whether the product evaluation model is successfully configured, and therefore obtain user information of multiple evaluated users. For the evaluated user, the terminal device records the user's historical evaluation level and the user parameter value of each native parameter of the historical user, that is, the aforementioned historical user parameter. The terminal device can judge whether the evaluation result is correct by comparing the actual output evaluation level and the historical evaluation level that has been judged to be legal, and determine the correct rate of the product evaluation model.

In S302, the variable value of each target variable is determined based on the historical user parameters, and each variable value is imported into the product evaluation model to calculate the training evaluation result of each evaluated user.

In this embodiment, the terminal device uses the user information of each evaluated user as a training sample, uses historical user parameters in the historical information as training input parameters, and sets the variable value of each target variable according to the historical user parameters, thereby passing the product evaluation model Calculate the variable value of each target variable, and output the training evaluation result of the evaluated user.

In S303, count the number of abnormalities in which the training evaluation results of the multiple evaluated users do not match the corresponding historical evaluation levels.

In this embodiment, the terminal device compares each training evaluation result with the historical evaluation level of the corresponding evaluated user. If the training evaluation result recognizes that the user level is consistent with the historical evaluation level, it will recognize that the training operation is not Abnormal; On the contrary, if the user level identified by the training evaluation result is inconsistent with the historical evaluation result, then the training operation abnormality is identified, so that the number of abnormal training operations can be counted as the number of abnormalities mentioned above.

In this embodiment, if the number of abnormalities is less than or equal to the preset abnormality threshold, it is identified that the product evaluation model belongs to a legal evaluation model, and the user to be evaluated can be rated by the legal evaluation model.

In S304, if the number of abnormalities is greater than a preset abnormality threshold, model abnormality information about the product evaluation model is generated.

In this embodiment, if the number of abnormalities detected is greater than the preset abnormality threshold, the correct rate of the identification product evaluation model is lower than the preset factory requirements. At this time, the product evaluation model needs to be adjusted, and the model abnormality information is output , So that the administrator can make manual adjustments.

In the embodiment of the application, the historical information of the identified user is used as a training sample to detect the accuracy of the generated product evaluation model, and generate model abnormal information for the product evaluation model with a lower accuracy rate, thereby improving the accuracy of the product evaluation model Sex.

FIG. 4 shows a specific implementation flowchart of a method S103 for generating a product evaluation model provided by the fourth embodiment of the present application. Referring to Fig. 4, with respect to the embodiment described in Figs. 1 to 3, the method S103 for generating a product evaluation model provided by this embodiment includes: S1031 to S1034, which are detailed as follows:

Further, the selecting the native variable matching the product keyword from the variable library and identifying the selected native variable as the target variable corresponding to the target product includes:

In S1031, the first degree of matching between the product keyword and the variable name of each native variable in the variable library is calculated.

In this embodiment, the terminal device determines whether the product keyword is the same as the variable name of each native variable in the variable library, recognizes the same number of characters between the product keyword and the variable name, and based on the same number of characters and variable The total characters of the name, calculate the first match between the two.

In S1032, if each of the first matching degrees is less than a preset matching threshold, obtain the synonymous keywords of the product keywords, and calculate the difference between the synonymous keywords and each of the variable names. The second degree of matching.

In this embodiment, if the terminal device detects that the first matching degree between the product keyword and each native variable is lower than the preset matching threshold, it will import the product keyword into the synonym through the synonym generation algorithm In the generation algorithm, the synonymous keywords of the product keywords are determined. Because users may use different phrases when writing product description information, such as the native variable "address", and the product keyword in the product description information is "location", and the physical meaning between the two is the same Therefore, it is possible to determine whether the variable library contains the native variable associated with the product keyword by calculating the second degree of matching between the synonymous keyword and the native variable. If there is a second degree of matching greater than the matching threshold, identify the native variable greater than the matching threshold as the target variable associated with the product keyword.

In S1033, if each of the second matching degrees is less than the matching threshold, obtain the variant keywords of the product keywords, and calculate the third match between the variant keywords and each of the variable names Degree; the variant keywords are keywords based on different languages from the product keywords.

In this embodiment, in addition to using different words for synonymous expression, it can also be expressed in different languages. Therefore, when it is determined that each native variable in the variable library does not match the product keyword and the same keyword, the terminal device Variant keywords of product keywords can be obtained. The terminal device can be configured with multiple preset languages to generate variant keywords for each preset language.

In S1034, if any one of the third matching degrees is greater than the matching threshold, identify the native variable corresponding to the third matching degree greater than the matching threshold as the target variable of the product keyword.

In this embodiment, since the variant keyword and the field keyword have the same physical meaning, if there is a variant keyword that matches the native field, the matched native field can be used as the native field for product keyword matching.

In the embodiment of the present application, the terminal device can extract the corresponding target variable from the variable library by acquiring multiple synonymous keywords and variant keywords of the product keyword, which improves the efficiency of acquiring the target variable.

FIG. 5 shows a specific implementation flow chart of a method for generating a product evaluation model provided by the fifth embodiment of the present application. Referring to Figure 5, with respect to the embodiment described in Figures 1 to 3, the method for generating a product evaluation model provided by this embodiment is to import the target variable and the derivative variable associated with the target variable into the The product evaluation template, after generating the product evaluation model of the target product, also includes: S501~S504, the details are as follows:

In S501, user information of the user to be evaluated is obtained, and the evaluation variable value of each target variable is determined according to the user information.

In this embodiment, after the terminal device generates the product evaluation model of the target product, it can determine the user evaluation level of the user who purchases the target product. The user to be evaluated is the user who needs to purchase the target product, and the terminal device is based on the user to be evaluated. Obtain user information and determine the evaluation variable value of each target variable from the user information.

In S502, a second derivative variable value of each derivative variable is calculated based on the evaluation variable value of each native variable.

In this embodiment, the terminal device imports the evaluation variable value of the target variable into the derivative variable conversion function according to the conversion function corresponding to each derivative variable to obtain the second derivative variable value for each derivative variable.

In S503, the evaluation variable value and the second derivative variable value are imported into the product evaluation template, and the user evaluation level of the user to be evaluated is calculated.

In this embodiment, the terminal device imports the calculated second derivative variable value and the evaluation variable value into the product evaluation model, and then the user evaluation level of the user to be evaluated can be calculated. If the user evaluation level is higher than or equal to the preset level threshold, the user is identified as a legitimate user, and the purchase operation is performed accordingly.

In S504, if the user evaluation level is lower than a preset level threshold, the user to be evaluated is identified as an abnormal user.

In this embodiment, if it is detected that the user evaluation level is lower than the level threshold, it means that the user does not have the authority to purchase the target product, and the user will be identified as an abnormal user.

In the embodiment of the present application, by determining the user evaluation level of the user to be evaluated, the legitimacy of the user is judged according to the user evaluation level, thereby improving the accuracy of user identification.

It should be understood that the size of the sequence number of each step in the foregoing embodiment does not mean the order of execution. The execution sequence of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiment of the present application.

FIG. 6 shows a structural block diagram of a device for generating a product evaluation model according to an embodiment of the present application. The device for generating a product evaluation model includes units for executing steps in the embodiment corresponding to FIG. 1. For details, please refer to the relevant description in the embodiment corresponding to FIG. 1 and FIG. 1. For ease of description, only the parts related to this embodiment are shown.

Referring to Figure 6, the device for generating the product evaluation model includes:

The derivative variable configuration unit 61 is configured to configure multiple derivative variables for each of the native variables in the variable library based on preset derivative variable conversion functions corresponding to different native variables, and establish the relationship between the native variables and the derivative variables Correspondence;

The product keyword acquisition unit 62 is configured to acquire product information of the target product, perform semantic analysis on the product information, and identify the product keywords contained in the product information;

The target variable identification unit 63 is configured to select the native variable matching the product keyword from the variable library, and identify the selected native variable as the target variable corresponding to the target product;

The derivative variable selection unit 64 is configured to obtain the derivative variables associated with each of the target variables based on the corresponding relationship;

The product evaluation model generating unit 65 is configured to download a product evaluation template of the product type according to the product type of the target product, and import the target variable and the derivative variable associated with the target variable into the product The evaluation template generates a product evaluation model of the target product.

Optionally, the derivative variable configuration unit 61 includes:

The in-use evaluation model analysis unit is used to analyze the created in-use evaluation model and determine the native variables and the derivative variables included in the in-use evaluation model;

The training variable value obtaining unit is configured to obtain the parameter value range of the native variable included in the in-use evaluation model, select multiple parameter nodes within the parameter value range, and configure training for each parameter node variable;

The first derivative variable value collection unit is configured to respectively import each of the training variable values into the in-use evaluation model, and collect the first derivative variable value output by the in-use evaluation model based on the training variable value;

A conversion function determining unit, configured to determine, based on the training variable value and the first derivative variable value, the derivative variable conversion function corresponding between the native variable and the derivative variable included in the in-use evaluation model .

Optionally, the device for generating the product evaluation model further includes:

The historical information obtaining unit is configured to obtain historical information of a plurality of evaluated users; the historical information includes historical user parameters and historical evaluation levels of the evaluated users;

A training evaluation result output unit, configured to determine the variable value of each target variable based on the historical user parameters, and import each variable value into the product evaluation model to calculate the training evaluation result of each of the evaluated users;

An abnormal number counting unit, configured to count the number of abnormalities whose training evaluation results of the multiple evaluated users do not match the corresponding historical evaluation level;

The abnormal model determining unit is configured to generate model abnormal information about the product evaluation model if the number of abnormalities is greater than a preset abnormal threshold.

Optionally, the target variable identification unit 63 includes:

A first matching degree calculation unit, configured to calculate a first matching degree between the product keyword and the variable name of each native variable in the variable library;

The second matching degree calculation unit is configured to obtain the synonymous keywords of the product keywords if each of the first matching degrees is less than a preset matching threshold, and calculate the synonymous keywords and each of the synonymous keywords. The second degree of match between variable names;

The third matching degree calculation unit is configured to, if each of the second matching degrees is less than the matching threshold, obtain the variant keywords of the product keywords, and calculate the difference between the variant keywords and each of the variable names The third degree of matching between; the variant keyword is a keyword based on a different language from the product keyword;

A target variable selecting unit, configured to identify the native variable corresponding to the third matching degree greater than the matching threshold as the target of the product keyword if any one of the third matching degrees is greater than the matching threshold variable.

The user information obtaining unit is configured to obtain user information of the user to be evaluated, and determine the evaluation variable value of each target variable according to the user information;

The evaluation variable value conversion unit is configured to calculate the second derivative variable value of each derivative variable based on the evaluation variable value of each native variable;

A user evaluation level calculation unit, configured to import the evaluation variable value and the second derivative variable value into the product evaluation template, and calculate the user evaluation level of the user to be evaluated;

The abnormal user identification unit is configured to identify the user to be evaluated as an abnormal user if the user evaluation level is lower than a preset level threshold.

Therefore, the device for generating the product evaluation model provided by the embodiment of the present application also does not require the user to manually configure the native variables, and configure the conversion function for each native variable, so as to obtain derivative variables, and can directly extract the target product-related items from the variable library. Variables can improve the generation efficiency of the evaluation model, and can also avoid the missing configuration of variables, which improves the accuracy of the product evaluation model.

Fig. 7 is a schematic diagram of a terminal device provided by another embodiment of the present application. As shown in FIG. 7, the terminal device 7 of this embodiment includes: a processor 70, a memory 71, and a computer program 72 stored in the memory 71 and running on the processor 70, such as the generation of a product evaluation model program. When the processor 70 executes the computer program 72, the steps in the embodiment of the method for generating each product evaluation model described above are implemented, such as S101 to S105 shown in FIG. 1. Alternatively, when the processor 70 executes the computer program 72, the functions of the units in the foregoing device embodiments, such as the functions of the modules 61 to 65 shown in FIG. 6, are realized.

Exemplarily, the computer program 72 may be divided into one or more units, and the one or more units are stored in the memory 71 and executed by the processor 70 to complete the application. The one or more units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program 72 in the terminal device 7. For example, the computer program 72 may be divided into a derivative variable configuration unit, a product keyword acquisition unit, a target variable identification unit, a derivative variable selection unit, and a product evaluation model generation unit. The specific functions of each unit are as described above.

The terminal device 7 may be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server. The terminal device may include, but is not limited to, a processor 70 and a memory 71. Those skilled in the art can understand that FIG. 7 is only an example of the terminal device 7 and does not constitute a limitation on the terminal device 7. It may include more or less components than shown in the figure, or a combination of certain components, or different components. For example, the terminal device may also include input and output devices, network access devices, buses, etc.

The so-called processor 70 may be a central processing unit (Central Processing Unit, CPU), other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), and application specific integrated circuits (Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor or the like.

The memory 71 may be an internal storage unit of the terminal device 7, such as a hard disk or a memory of the terminal device 7. The memory 71 may also be an external storage device of the terminal device 7, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), or a secure digital (Secure Digital, SD) equipped on the terminal device 7. Card, Flash Card, etc. Further, the memory 71 may also include both an internal storage unit of the terminal device 7 and an external storage device. The memory 71 is used to store the computer program and other programs and data required by the terminal device. The memory 71 can also be used to temporarily store data that has been output or will be output.

In addition, the functional units in each embodiment of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or software functional unit.

The above-mentioned embodiments are only used to illustrate the technical solutions of the present application, not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that it can still implement the foregoing The technical solutions recorded in the examples are modified, or some of the technical features are equivalently replaced; these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the application, and should be included in Within the scope of protection of this application.

Claims

A method for generating product evaluation models, including:

Based on the preset derivative variable conversion functions corresponding to different native variables, configure multiple derivative variables for each native variable in the variable library, and establish the corresponding relationship between the native variable and the derivative variable;

Obtain product information of the target product, perform semantic analysis on the product information, and identify product keywords contained in the product information;

Selecting the native variable matching the product keyword from the variable library, and identifying the selected native variable as the target variable corresponding to the target product;

Obtaining the derivative variable associated with each of the target variables based on the corresponding relationship;

According to the product type of the target product, download the product evaluation template of the product type, and import the target variable and the derivative variable associated with the target variable into the product evaluation template to generate the evaluation template of the target product Product evaluation model.
The generating method according to claim 1, wherein the preset derivative variable conversion function corresponding to different native variables respectively configures multiple derivative variables for each native variable in the variable library, and establishes the native variable Before the corresponding relationship with the derivative variable, it also includes:

Analyze the created in-use evaluation model, and determine the native variables and the derivative variables included in the in-use evaluation model;

Acquiring the parameter value range of the native variable included in the in-use evaluation model, selecting a plurality of parameter nodes within the parameter value range, and configuring a training variable value for each parameter node;

Import each of the training variable values into the in-use evaluation model, and collect the first derivative variable values output by the in-use evaluation model based on the training variable values;

According to the training variable value and the first derivative variable value, the derivative variable conversion function between the native variable and the derivative variable included in the in-use evaluation model is determined.
4. The generating method according to claim 1, wherein after the target variable and the derivative variable associated with the target variable are imported into the product evaluation template to generate a product evaluation model of the target product, Also includes:

Acquiring historical information of multiple evaluated users; the historical information includes historical user parameters and historical evaluation levels about the evaluated users;

Determine the variable value of each of the target variables based on the historical user parameters, and import each of the variable values into the product evaluation model, and calculate the training evaluation result of each of the evaluated users;

Counting the number of abnormalities in which the training evaluation results of the multiple evaluated users do not match the corresponding historical evaluation levels;

If the number of abnormalities is greater than a preset abnormality threshold, model abnormality information about the product evaluation model is generated.
The generating method according to any one of claims 1 to 3, wherein the selecting the native variable matching the product keyword from the variable library, and identifying the selected native variable as the selected native variable Describe the target variables corresponding to the target product, including:

Calculating the first degree of matching between the product keyword and the variable name of each native variable in the variable library;

If each of the first matching degrees is less than the preset matching threshold, obtain the synonymous keywords of the product keywords, and calculate the second matching degrees between the synonymous keywords and each of the variable names ；

If each of the second matching degrees is less than the matching threshold, acquiring the variant keywords of the product keywords, and calculating the third matching degrees between the variant keywords and each of the variable names; Variant keywords are keywords based on different languages from the product keywords;

If any one of the third matching degrees is greater than the matching threshold, identifying the native variable corresponding to the third matching degree greater than the matching threshold as the target variable of the product keyword.
The generating method according to claim 1, wherein after importing the target variable and the derivative variable associated with the target variable into the product evaluation template to generate a product evaluation model of the target product, the method further comprises :

Acquiring user information of the user to be evaluated, and determining the evaluation variable value of each of the target variables according to the user information;

Calculating the second derivative variable value of each derivative variable based on the evaluation variable value of each of the primary variables;

Importing the evaluation variable value and the second derivative variable value into the product evaluation template, and calculating the user evaluation level of the user to be evaluated;

If the user evaluation level is lower than a preset level threshold, the user to be evaluated is identified as an abnormal user.
The generating method according to claim 2, wherein after importing the target variable and the derivative variable associated with the target variable into the product evaluation template to generate the product evaluation model of the target product, the method further comprises :

Acquiring user information of the user to be evaluated, and determining the evaluation variable value of each of the target variables according to the user information;

Calculating the second derivative variable value of each derivative variable based on the evaluation variable value of each of the primary variables;

Importing the evaluation variable value and the second derivative variable value into the product evaluation template, and calculating the user evaluation level of the user to be evaluated;

If the user evaluation level is lower than a preset level threshold, the user to be evaluated is identified as an abnormal user.
4. The generating method according to claim 3, wherein after importing the target variable and the derivative variable associated with the target variable into the product evaluation template to generate a product evaluation model of the target product, the method further comprises :

Acquiring user information of the user to be evaluated, and determining the evaluation variable value of each of the target variables according to the user information;

Calculating the second derivative variable value of each derivative variable based on the evaluation variable value of each of the primary variables;

Importing the evaluation variable value and the second derivative variable value into the product evaluation template, and calculating the user evaluation level of the user to be evaluated;

If the user evaluation level is lower than a preset level threshold, the user to be evaluated is identified as an abnormal user.
A terminal device, wherein the terminal device includes a memory, a processor, and a computer program that is stored in the memory and can run on the processor, and the processor executes the computer program as follows:

Based on the preset derivative variable conversion functions corresponding to different native variables, configure multiple derivative variables for each native variable in the variable library, and establish the corresponding relationship between the native variable and the derivative variable;

Obtain product information of the target product, perform semantic analysis on the product information, and identify product keywords contained in the product information;

Selecting the native variable matching the product keyword from the variable library, and identifying the selected native variable as the target variable corresponding to the target product;

Obtaining the derivative variable associated with each of the target variables based on the corresponding relationship;

According to the product type of the target product, download the product evaluation template of the product type, and import the target variable and the derivative variable associated with the target variable into the product evaluation template to generate the evaluation template of the target product Product evaluation model.
The terminal device according to claim 8, wherein the preset derivative variable conversion function corresponding to different native variables respectively configures multiple derivative variables for each native variable in the variable library, and establishes the native variable Before the corresponding relationship with the derivative variable, it also includes:

Analyze the created in-use evaluation model, and determine the native variables and the derivative variables included in the in-use evaluation model;

Acquiring the parameter value range of the native variable included in the in-use evaluation model, selecting a plurality of parameter nodes within the parameter value range, and configuring a training variable value for each parameter node;

Import each of the training variable values into the in-use evaluation model, and collect the first derivative variable values output by the in-use evaluation model based on the training variable values;

According to the training variable value and the first derivative variable value, the derivative variable conversion function between the native variable and the derivative variable included in the in-use evaluation model is determined.
8. The terminal device according to claim 8, wherein after the target variable and the derivative variable associated with the target variable are imported into the product evaluation template to generate a product evaluation model of the target product, Also includes:

Acquiring historical information of multiple evaluated users; the historical information includes historical user parameters and historical evaluation levels about the evaluated users;

Determine the variable value of each of the target variables based on the historical user parameters, and import each of the variable values into the product evaluation model, and calculate the training evaluation result of each of the evaluated users;

Counting the number of abnormalities in which the training evaluation results of the multiple evaluated users do not match the corresponding historical evaluation levels;

If the number of abnormalities is greater than a preset abnormality threshold, model abnormality information about the product evaluation model is generated.
The terminal device according to any one of claims 8-10, wherein the selecting the native variable matching the product keyword from the variable library, and identifying the selected native variable as the Describe the target variables corresponding to the target product, including:

Calculating the first degree of matching between the product keyword and the variable name of each native variable in the variable library;

If each of the first matching degrees is less than the preset matching threshold, obtain the synonymous keywords of the product keywords, and calculate the second matching degrees between the synonymous keywords and each of the variable names ；

If each of the second matching degrees is less than the matching threshold, acquiring the variant keywords of the product keywords, and calculating the third matching degrees between the variant keywords and each of the variable names; Variant keywords are keywords based on different languages from the product keywords;

If any one of the third matching degrees is greater than the matching threshold, identifying the native variable corresponding to the third matching degree greater than the matching threshold as the target variable of the product keyword.
The terminal device according to claim 8, wherein, after importing the target variable and the derivative variable associated with the target variable into the product evaluation template to generate a product evaluation model of the target product, further comprising: :

Acquiring user information of the user to be evaluated, and determining the evaluation variable value of each of the target variables according to the user information;

Calculating the second derivative variable value of each derivative variable based on the evaluation variable value of each of the primary variables;

Importing the evaluation variable value and the second derivative variable value into the product evaluation template, and calculating the user evaluation level of the user to be evaluated;

If the user evaluation level is lower than a preset level threshold, the user to be evaluated is identified as an abnormal user.
The terminal device according to claim 9, wherein after importing the target variable and the derivative variable associated with the target variable into the product evaluation template to generate a product evaluation model of the target product, further comprising: :

Acquiring user information of the user to be evaluated, and determining the evaluation variable value of each of the target variables according to the user information;

Calculating the second derivative variable value of each derivative variable based on the evaluation variable value of each of the primary variables;

Importing the evaluation variable value and the second derivative variable value into the product evaluation template, and calculating the user evaluation level of the user to be evaluated;

If the user evaluation level is lower than a preset level threshold, the user to be evaluated is identified as an abnormal user.
The terminal device according to claim 10, wherein, after importing the target variable and the derivative variable associated with the target variable into the product evaluation template to generate a product evaluation model of the target product, further comprising: :

Acquiring user information of the user to be evaluated, and determining the evaluation variable value of each of the target variables according to the user information;

Calculating the second derivative variable value of each derivative variable based on the evaluation variable value of each of the primary variables;

Importing the evaluation variable value and the second derivative variable value into the product evaluation template, and calculating the user evaluation level of the user to be evaluated;

If the user evaluation level is lower than a preset level threshold, the user to be evaluated is identified as an abnormal user.
A computer-readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the following steps:

Based on the preset derivative variable conversion functions corresponding to different native variables, configure multiple derivative variables for each native variable in the variable library, and establish the corresponding relationship between the native variable and the derivative variable;

Obtain product information of the target product, perform semantic analysis on the product information, and identify product keywords contained in the product information;

Selecting the native variable matching the product keyword from the variable library, and identifying the selected native variable as the target variable corresponding to the target product;

Obtaining the derivative variable associated with each of the target variables based on the corresponding relationship;

According to the product type of the target product, download the product evaluation template of the product type, and import the target variable and the derivative variable associated with the target variable into the product evaluation template to generate the evaluation template of the target product Product evaluation model.
The computer-readable storage medium according to claim 15, wherein the preset derivative variable conversion functions corresponding to different native variables respectively configure multiple derivative variables for each native variable in the variable library, and establish all Before describing the corresponding relationship between the native variable and the derived variable, it also includes:

Analyze the created in-use evaluation model, and determine the native variables and the derivative variables included in the in-use evaluation model;

Acquiring the parameter value range of the native variable included in the in-use evaluation model, selecting a plurality of parameter nodes within the parameter value range, and configuring a training variable value for each parameter node;

Import each of the training variable values into the in-use evaluation model, and collect the first derivative variable values output by the in-use evaluation model based on the training variable values;

According to the training variable value and the first derivative variable value, the derivative variable conversion function between the native variable and the derivative variable included in the in-use evaluation model is determined.
15. The computer-readable storage medium according to claim 15, wherein the target variable and the derivative variable associated with the target variable are imported into the product evaluation template to generate a product evaluation of the target product After the model, it also includes:

Acquiring historical information of multiple evaluated users; the historical information includes historical user parameters and historical evaluation levels about the evaluated users;

Determine the variable value of each of the target variables based on the historical user parameters, and import each of the variable values into the product evaluation model, and calculate the training evaluation result of each of the evaluated users;

Counting the number of abnormalities in which the training evaluation results of the multiple evaluated users do not match the corresponding historical evaluation levels;

If the number of abnormalities is greater than a preset abnormality threshold, model abnormality information about the product evaluation model is generated.
18. The computer-readable storage medium according to any one of claims 15-17, wherein the selected native variable matching the product keyword is selected from the variable library, and the selected native variable The target variable identified as the corresponding target product includes:

Calculating the first degree of matching between the product keyword and the variable name of each native variable in the variable library;

If each of the first matching degrees is less than the preset matching threshold, obtain the synonymous keywords of the product keywords, and calculate the second matching degrees between the synonymous keywords and each of the variable names ；

If each of the second matching degrees is less than the matching threshold, acquiring the variant keywords of the product keywords, and calculating the third matching degrees between the variant keywords and each of the variable names; Variant keywords are keywords based on different languages from the product keywords;

If any one of the third matching degrees is greater than the matching threshold, identifying the native variable corresponding to the third matching degree greater than the matching threshold as the target variable of the product keyword.
The computer-readable storage medium according to claim 15, wherein, after the target variable and the derivative variable associated with the target variable are imported into the product evaluation template, a product evaluation model of the target product is generated ,Also includes:

Acquiring user information of the user to be evaluated, and determining the evaluation variable value of each of the target variables according to the user information;

Calculating the second derivative variable value of each derivative variable based on the evaluation variable value of each of the primary variables;

Importing the evaluation variable value and the second derivative variable value into the product evaluation template, and calculating the user evaluation level of the user to be evaluated;

If the user evaluation level is lower than a preset level threshold, the user to be evaluated is identified as an abnormal user.
The computer-readable storage medium according to claim 16 or 17, wherein the target variable and the derivative variable associated with the target variable are imported into the product evaluation template to generate a product evaluation of the target product After the model, it also includes:

Acquiring user information of the user to be evaluated, and determining the evaluation variable value of each of the target variables according to the user information;

Calculating the second derivative variable value of each derivative variable based on the evaluation variable value of each of the primary variables;

Importing the evaluation variable value and the second derivative variable value into the product evaluation template, and calculating the user evaluation level of the user to be evaluated;

If the user evaluation level is lower than a preset level threshold, the user to be evaluated is identified as an abnormal user.