WO2020073727A1

WO2020073727A1 - Risk forecast method, device, computer apparatus, and storage medium

Info

Publication number: WO2020073727A1
Application number: PCT/CN2019/099382
Authority: WO
Inventors: 于修铭; 汪伟; 肖京
Original assignee: 平安科技（深圳）有限公司
Priority date: 2018-10-11
Filing date: 2019-08-06
Publication date: 2020-04-16
Also published as: CN109523117A

Abstract

A risk forecast method comprises: receiving a forecast request sent by a terminal, and acquiring information of each dimension of a company corresponding to the forecast request carrying a company identifier, the company identifier corresponding to the information of each dimension of the company; extracting a company index from the information of each dimension of the company, the company index comprising various financial indexes, legal proceeding information, public sentiment information, and import and export lists; inputting the company index as a forecasting feature into an available forecasting model obtained from previous training, and outputting a forecast label corresponding to the company index, wherein the available forecasting model is obtained from training according to a training data set and indication information corresponding to sample data, and the forecast label comprises a risk-involvement forecast label and a risk-free forecast label; acquiring forecast information corresponding to the company index according to the forecast label, the forecast information comprising risk-involvement forecast information corresponding to the risk-involvement forecast label and risk-free forecast information corresponding to the risk-free forecast label; and sending the forecast information to a corresponding terminal.

Description

Risk prediction method, device, computer equipment and storage medium

Cross-reference of related applications

This application requires priority to be submitted to the Chinese Patent Office on October 11, 2018, with the application number 2018111836725, and the priority of the Chinese patent application titled "risk prediction methods, devices, computer equipment and storage media", the entire contents of which are incorporated by reference In this application.

Technical field

This application relates to a risk prediction method, device, computer equipment and storage medium.

Background technique

The risk early warning system is based on the characteristics of the research object, by collecting relevant data and information, monitoring the changing trend of risk factors, and evaluating the strength of various risk states deviating from the early warning line, sending early warning signals to the decision-making layer and taking pre-control in advance Countermeasure system. Therefore, to build an early warning system, you must first build an evaluation index system and analyze and process the index categories. Second, based on the early warning model, comprehensive evaluation of the evaluation index system. Finally, the early warning interval is set according to the judgment results, and corresponding countermeasures are taken.

However, the inventor realized that most of the traditional enterprise risk early warning systems are built based on statistical methods or according to expert scorecards. The key part is the pre-calculated thresholds of indicators, the validity of indicators, and the weight of intermediate indicators. These data need to be calculated manually, and errors are likely to occur in the calculation process and affect the risk prediction effect.

Summary of the invention

According to various embodiments disclosed in the present application, a risk prediction method, apparatus, computer equipment, and storage medium are provided.

A risk prediction method includes:

Receiving a prediction request sent by a terminal, and acquiring information of each dimension of the enterprise corresponding to the prediction request; the prediction request carries an enterprise identifier, and the enterprise identifier corresponds to each dimension information of the enterprise;

Extract enterprise indicators from all dimensions of the enterprise; the enterprise indicators include various financial indicators, legal litigation information, public opinion information, and import and export lists;

Input the enterprise index as a prediction feature into an available prediction model obtained by pre-training, and output a prediction label corresponding to the enterprise index; the available prediction model is trained based on the prompt information corresponding to the training data set and sample data; Forecast labels include risk prediction labels and risk-free prediction labels;

Obtaining prediction information corresponding to the enterprise indicator according to the prediction tag; the prediction information includes risk prediction information corresponding to the risk prediction tag, and risk-free prediction information corresponding to the risk-free prediction tag; and

Send the prediction information to the corresponding terminal.

A risk prediction device includes:

The dimension information obtaining module is used to receive the prediction request sent by the terminal, and obtain each dimension information of the enterprise corresponding to the prediction request; the prediction request carries an enterprise identifier, and the enterprise identifier corresponds to each dimension information of the enterprise ;

The enterprise index extraction module is used to extract enterprise indexes from all dimensions of the enterprise; the enterprise indexes include various financial indexes, legal litigation information, public opinion information, and import and export lists;

The tag acquisition module is used to input the enterprise index as a prediction feature into an available prediction model that is pre-trained, and output a prediction tag corresponding to the enterprise index; the available prediction model is based on a prompt corresponding to the training data set and sample data Information training; the prediction labels include risk prediction labels and risk-free prediction labels;

The prediction information obtaining module is used to obtain prediction information corresponding to the enterprise index according to the prediction label; the prediction information includes risk prediction information corresponding to the risk prediction label, and corresponding to the risk-free prediction label No risk prediction information; and

The sending module is used to send the prediction information to the corresponding terminal.

A computer device includes a memory and one or more processors. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the one or more processors are executed The following steps:

Send the prediction information to the corresponding terminal.

One or more non-volatile storage media storing computer readable instructions, which when executed by one or more processors, cause the one or more processors to perform the following steps:

Send the prediction information to the corresponding terminal.

The details of one or more embodiments of the application are set forth in the drawings and description below. Other features and advantages of this application will become apparent from the description, drawings, and claims.

BRIEF DESCRIPTION

In order to more clearly explain the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. Those of ordinary skill in the art can obtain other drawings based on these drawings without creative efforts.

FIG. 1 is an application scenario diagram of a risk prediction method according to one or more embodiments.

FIG. 2 is a schematic flowchart of a risk prediction method according to one or more embodiments.

FIG. 3 is a schematic flowchart of constructing an initial model according to one or more embodiments.

4 is a schematic flowchart of using a sample label as a prediction label of an initial model according to one or more embodiments.

FIG. 5 is a block diagram of a risk prediction device according to one or more embodiments.

Figure 6 is a block diagram of a computer device in accordance with one or more embodiments.

detailed description

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

The risk prediction method provided by this application can be applied in the application environment shown in FIG. 1. The terminal 102 and the server 104 communicate via the network. The server 104 receives the prediction request sent by the terminal 102 and obtains the information of each dimension of the enterprise corresponding to the prediction request. The prediction request carries the enterprise logo, and the enterprise logo corresponds to the information of each dimension of the enterprise. The server 104 extracts enterprise indicators from various dimensions of the enterprise. The enterprise indicators include various financial indicators, legal litigation information, public opinion information, and import and export lists. The enterprise index is used as a prediction feature to input into the available prediction model that is pre-trained, and the prediction label corresponding to the enterprise index is output. The prediction model can be trained according to the prompt information corresponding to the training data set and sample data. Risk prediction label. The server 104 obtains the prediction information corresponding to the enterprise index according to the prediction tag, the prediction information includes the risk prediction information corresponding to the risk prediction tag, and the risk-free prediction information corresponding to the risk-free prediction tag, and sends the prediction information to the corresponding terminal 102. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices. The server 104 may be implemented by an independent server or a server cluster composed of multiple servers.

In one of the embodiments, as shown in FIG. 2, a risk prediction method is provided. Taking the method applied to the server in FIG. 1 as an example for illustration, it includes the following steps:

S202. The server receives the prediction request sent by the terminal, and obtains information on each dimension of the enterprise corresponding to the prediction request. The prediction request carries an enterprise ID, and the enterprise ID corresponds to each dimension information of the enterprise.

Among them, the prediction request carries the enterprise logo, and there is a correspondence between the enterprise logo and each dimension information of the enterprise, and the server can obtain each dimension information of the enterprise corresponding to the enterprise logo according to the correspondence between the enterprise logo and each dimension information of the enterprise .

Specifically, the server receives the prediction request sent by the terminal, and parses the prediction request to obtain the enterprise identification carried in the prediction request. Obtain the correspondence between the enterprise logo and each dimension of the enterprise, and according to the correspondence between the acquired enterprise logo and each dimension of the enterprise, obtain the dimension information of the enterprise corresponding to the enterprise logo from the database.

Further, in the actual application scenario, the enterprise management system performs daily operations, and can call the interface to obtain information and data related to the enterprise from other platforms connected to the enterprise management system, based on the acquired enterprise-related information And data to generate information about each dimension of the corresponding enterprise. After receiving the prediction request sent by the corresponding terminal where the enterprise is located, the server parses the prediction request sent by the corresponding terminal and obtains the enterprise identification carried in the prediction request. Furthermore, the server can identify the corresponding enterprise according to the enterprise identification, and obtain the information of each dimension of the enterprise corresponding to the enterprise identification from the database.

S204. The server extracts enterprise indicators from all dimensions of the enterprise. The enterprise indicators include various financial indicators, legal litigation information, public opinion information, and import and export lists.

Specifically, the server acquires the information of each dimension of the enterprise from the database, and extracts the enterprise index from the information of each dimension of the enterprise. Among them, there is a correspondence between each dimension information of the enterprise and the enterprise index. The server can obtain the correspondence between each dimension information of the enterprise and the enterprise index, and according to the correspondence between the dimension information of the enterprise and the enterprise index, Separately extract the enterprise indexes corresponding to the information of each dimension.

Enterprise indicators include various financial indicators, legal litigation information, public opinion information, and import and export lists. Among them, financial indicators refer to the relative indicators for enterprises to summarize and evaluate the financial status and operating results, including: debt solvency indicators, including asset-liability ratio, current ratio, quick ratio; operating ability indicators, including accounts receivable turnover , Inventory turnover rate; profitability indicators, including capital profit rate, sales profit rate (operating income tax rate), cost expense profit rate, etc.

The legal litigation information indicates the legal situation of the enterprise's participation in legal affairs. Public opinion is the abbreviation of "public opinion situation", which refers to the occurrence, development and change of intermediary social events in a certain social space, and the public as the subject of social managers, enterprises, individuals and other various organizations as the object Social attitudes generated and held by their political, social, and moral orientations, including positive and negative public opinion, can also express the beliefs, attitudes, opinions, and emotions expressed by more people about various phenomena and problems in society Wait for the sum of performance. The import and export list shows the import and export situation of the company's products, including product type, corresponding product quantity, product selling price and purchase price, etc.

S206, the server inputs the enterprise index as a prediction feature into the available prediction model that is pre-trained, and outputs the prediction label corresponding to the enterprise index. The available prediction model is trained based on the prompt information corresponding to the training data set and sample data. Labels and risk-free predictive labels.

Specifically, the server inputs the enterprise index as a prediction feature into the available prediction model, and obtains the risk prediction label and the risk-free prediction label corresponding to the enterprise index, respectively. Furthermore, the server can obtain the risk prediction information corresponding to the risk prediction label according to the correspondence between the preset risk prediction label and the risk prediction information, and according to the correspondence between the preset risk-free prediction label and the risk-free prediction information To obtain the risk-free prediction information corresponding to the risk-free prediction label.

Further, the server obtains the sample data from the database and extracts the sample label from the sample data, and uses the sample label as the prediction label of the initial model. At the same time, the server obtains the indexes corresponding to the sample data and uses the indexes corresponding to the sample data as the prediction features of the initial model, and then generates a training data set according to the prediction features and the prediction labels. Use the prompt information corresponding to the training data set and the sample data to train the initial model to obtain an available prediction model.

The initial model is an existing risk prediction model, but it is not applicable to all companies that need to make risk predictions. It is necessary to use the sample label, that is, the prediction label, and the prediction feature, that is, the index corresponding to the sample data to generate the training data set. The server can use the prompt information corresponding to the training data set and the sample data to train the initial model, and obtain the available test model corresponding to the enterprise according to the initial model after the training.

S208. The server obtains prediction information corresponding to the enterprise index according to the prediction tag. The prediction information includes risk prediction information corresponding to the risk prediction tag and risk-free prediction information corresponding to the risk-free prediction tag.

Specifically, the server obtains the risk prediction information corresponding to the risk prediction tag according to the correspondence between the preset risk prediction tag and the risk prediction information, and according to the correspondence between the preset risk-free prediction tag and the risk-free prediction information Relationship to obtain risk-free prediction information corresponding to the risk-free prediction label.

The risk prediction information indicates that the corresponding company has risk information, and the prediction information received by the corresponding company from the server is risk prediction information, which needs to be further investigated and processed based on the received risk prediction information. The risk-free forecast information indicates that there is no risk information for the corresponding enterprise, and the forecast information sent by the server received by the corresponding enterprise is the risk-free forecast information.

S210: The server sends the prediction information to the corresponding terminal.

In the above risk prediction method, the server obtains the information of each dimension of the enterprise according to the prediction request sent by the terminal, and extracts the enterprise index from the information of each dimension of the enterprise. Therefore, the enterprise index can be input into the available prediction model that is pre-trained as the prediction feature, and the prediction information corresponding to the enterprise index can be obtained, and the prediction information can be sent to the terminal. In the case of data update, there is no need to repeatedly perform data preprocessing on each indicator, which can reduce resource consumption, and at the same time, send corresponding prediction information to corresponding terminals for different prediction tags, which can further improve the risk prediction effect.

In one of the embodiments, as shown in FIG. 3, the steps of constructing the initial model include:

S302. The server obtains sample data from the database, and extracts sample labels from the sample data, and uses the sample labels as the prediction labels of the initial model.

Specifically, the server extracts sample tags from the sample data, and obtains the attributes of the sample tags, and divides the sample data into the first type of samples and the second type of samples according to the attributes. The server respectively obtains a first sample label corresponding to the first type of sample and a second sample label corresponding to the second type of sample. The server uses the first sample label as the risk prediction label of the initial model, and uses the second sample label as the risk-free prediction label of the initial model.

The attribute of the sample label is used to indicate whether the sample data corresponding to the sample label carries risk data. The server may divide the sample data into a first type sample and a second type sample according to the attributes of the sample label. The first type sample is sample data carrying risk data, and the second type sample is sample data not carrying risk data. Furthermore, the first sample label is a sample label carrying risk data, and the second sample label is a sample label without risk data.

S304. The server acquires the sample index corresponding to the sample data, and uses the sample index as the prediction feature of the initial model.

Specifically, the server obtains the index corresponding to the sample data by acquiring the correspondence between the sample data and the index, and according to the correspondence between the sample data and the index. The server obtains the correlation between the sample data and the predicted features of the initial model, and uses the indexes corresponding to the sample data as the predicted features of the initial model according to the correlation between the sample data and the predicted features.

Further, the server obtains the sample data set from the database and stores the sample data set in the preset object. Furthermore, the server acquires the training parameters corresponding to the training function by calling the training function in the database and according to the correspondence between the preset training function and the training parameter. The sample data set is called from the preset object, and the initial model is constructed according to the training parameters and the sample data set.

The training parameters include the maximum depth of the tree, the contraction step size and the number of iterations. The maximum depth of the tree represents the longest path from the root node to the leaf node plus 1, and from a recursive point of view, the depth of the tree is equal to the depth of its largest left and right subtree plus 1. The contraction step length means that a value is added to each contraction operation by adding a certain number (this is the step length), and this operation is repeatedly performed. Iteration refers to the process of repeatedly performing a series of calculation steps and sequentially determining the subsequent amounts from the previous amount. Each result of this process is obtained by performing the same calculation step on the previous result, and the number of iterations means The number of times to repeat a series of calculation steps.

S306. The server generates a training data set according to the predicted feature and the predicted label.

Specifically, the prediction feature is an indicator corresponding to the sample data, and the prediction label is a sample label of the sample data, including a first sample label and a second sample label, where the first sample label is a sample label carrying risk data, the second The sample label is a sample label of non-risk data. The server generates a training data set according to the index corresponding to the sample data and the first sample label, and the index corresponding to the sample data and the second sample label.

S308. The server uses the prompt information corresponding to the training data set and the sample data to train the initial model to obtain an available prediction model.

Specifically, the server trains the sample data set according to the training parameters to obtain the training data set, and applies the prompt information corresponding to the training data set and the sample data to the initial model to obtain the prediction percentage. According to the prediction percentage and sample data set, build a usable prediction model.

The training parameters include the maximum depth of the tree, the contraction step size and the number of iterations. The maximum depth of the tree indicates that the longest path from the root node to the leaf node is increased by 1, and the contraction step indicates that a value is added to each contraction operation, plus a certain number (this is the step size), and repeated execution In this operation, the number of iterations indicates the number of times to repeat a series of calculation steps.

The initial model is an existing risk prediction model, but it is not suitable for all companies that need to make risk predictions. The prompt information corresponding to the training data set and sample data is applied to the initial model to obtain the prediction percentage. The prediction percentage is used to represent the prompt information corresponding to the training data set and the sample data. After being applied to the initial model, it is obtained by calculating the degree of association between the training data set and the sample data.

In the above risk prediction method, the server extracts the sample label from the obtained sample data, and uses the sample label as the prediction label of the initial model. Obtain the index corresponding to the sample data, and use the index corresponding to the sample data as the prediction feature of the initial model, and generate the training data set according to the prediction feature and the prediction label, and then use the prompt information corresponding to the training data set and the sample data to carry out the initial model Train to get a usable prediction model. Therefore, by directly using the sample label as the prediction label of the initial model, it is possible to reduce the need for reprocessing the sample data every time the data is updated, and reduce resource consumption.

In one of the embodiments, as shown in FIG. 4, the step of extracting the sample label from the sample data and using the sample label as the prediction label of the initial model includes:

S402. The server extracts sample tags from the sample data and obtains the attributes of the sample tags.

S404. The server divides the sample data into the first type of samples and the second type of samples according to the attributes.

Specifically, the server obtains the sample data and the correspondence between the sample data and the sample label, and extracts the corresponding sample label from the sample data according to the correspondence between the sample data and the sample label. The attribute of the sample label is used to indicate whether the sample data corresponding to the sample label carries risk data. The server may divide the sample data into a first type sample and a second type sample according to the attributes of the sample label. The first type sample is sample data carrying risk data, and the second type sample is sample data not carrying risk data.

S406. The server obtains a first sample label corresponding to the first type of sample and a second sample label corresponding to the second type of sample.

Specifically, the server acquires the corresponding relationship between the first type of sample and the first sample label, and obtains the first corresponding to the first type of sample according to the corresponding relationship between the first type of sample and the first sample label Sample label. The first sample label is a sample label that carries risk data. The server obtains the second sample label corresponding to the second type sample by acquiring the corresponding relationship between the second type sample and the second sample label, and according to the corresponding relationship between the second type sample and the second sample label. Among them, the second sample label is a sample label of non-risk data.

S408. The server uses the first sample label as the risk prediction label of the initial model, and uses the second sample label as the risk-free prediction label of the initial model.

Specifically, the first sample label is a sample label carrying risk data, and the second sample label is a sample label without risk data. The server uses the first sample label, that is, the sample label that carries risk data, as the risk prediction label of the initial model, and corresponds to the first type of sample that carries risk data. The second sample label, that is, the sample label that does not carry risk data, is used as the risk-free prediction label of the initial model, and corresponds to the second type sample that does not carry risk data.

In the above step of using the sample label as the prediction label of the initial model, the server extracts the sample label from the sample data and obtains the attribute of the sample label, and divides the sample data into the first type sample and the second type sample according to the attribute. Obtain the first sample label corresponding to the first type sample and the second sample label corresponding to the second type sample, and use the first sample label as the risk prediction label of the initial model and the second sample label as the initial model Of risk-free prediction labels. Therefore, the sample tags carrying the risk data and the non-risk data can be distinguished, which is beneficial for performing targeted sample data processing and improving work efficiency.

In one of the embodiments, a risk prediction method is provided, and the method further includes the following steps:

The server obtains the sample data set from the database and stores the sample data set in the preset object; calls the training function in the database; according to the correspondence between the preset training function and the training parameter, obtains the training corresponding to the training function Parameters; call the sample data set from the preset object, and build the initial model based on the training parameters and the sample data set.

Specifically, the server obtains the training parameters corresponding to the called training function by acquiring the correspondence between the training function and the training parameter, and according to the correspondence between the training function and the training parameter. Use the training parameters to train the sample data set called from the preset object to obtain the initial model.

In the above risk prediction method, the server obtains the training parameters corresponding to the training function according to the correspondence between the preset training function and the training parameter, and calls the sample data set from the preset object, according to the training parameter and the sample data set To build the initial model. Therefore, the establishment of the initial model can be achieved, and the training data can be used to train the sample data set, thereby providing a corresponding basis for the subsequent available prediction models, and improving work efficiency.

In one of the embodiments, the step of receiving the prediction request sent by the terminal and obtaining the information of each dimension of the enterprise according to the prediction request includes:

The server receives and parses the prediction request sent by the terminal; obtains the enterprise logo carried in the prediction request; based on the enterprise logo

Correspondence between the identification and the information of each dimension of the enterprise to obtain the information of each dimension of the enterprise corresponding to the enterprise logo.

In the above step of obtaining the information of each dimension of the enterprise according to the prediction request, the server receives and parses the prediction request sent by the terminal, obtains the request tag carried by the prediction request, and obtains the dimension information of the enterprise corresponding to the request tag. Therefore, it is possible to obtain the information of each dimension of the targeted enterprise, clarify the correspondence between different prediction requests and the corresponding enterprise, and facilitate the acquisition and storage of the dimension information of the personalized enterprise.

In one of the embodiments, the step of obtaining prediction information corresponding to the enterprise index according to the prediction label includes:

The server obtains the risk prediction information corresponding to the risk prediction tag according to the correspondence between the preset risk prediction tag and the risk prediction information; according to the correspondence between the preset risk-free prediction tag and the risk-free prediction information, the Risk-free prediction information corresponding to the risk-free prediction label.

Specifically, the risk prediction information indicates that there is risk information for the corresponding enterprise, and the prediction information received by the corresponding enterprise from the server is risk prediction information, which needs to be further investigated and processed according to the received risk prediction information. The risk-free forecast information indicates that there is no risk information for the corresponding enterprise, and the forecast information sent by the server received by the corresponding enterprise is the risk-free forecast information.

In the above steps, the server can obtain the risk prediction information corresponding to the risk prediction label and the risk-free prediction information corresponding to the risk-free prediction label by separately acquiring the risk prediction label and the risk-free prediction label corresponding to the enterprise index. Therefore, corresponding prediction labels are obtained for different enterprise indicators, and corresponding risk prediction information and risk-free prediction information can be obtained separately, thereby improving the accuracy of risk prediction.

In one of the embodiments, using the prompt information corresponding to the training data set and the sample data to train the initial model to obtain a usable prediction model includes:

The server trains the sample data set according to the training parameters to obtain the training data set; applies the prompt information corresponding to the training data set and the sample data to the initial model to obtain the prediction percentage; according to the prediction percentage and the sample data set, constructs an available prediction model.

Specifically, the initial model is an existing risk prediction model, but it is not applicable to all enterprises that need to make risk predictions. The prompt information corresponding to the training data set and sample data is applied to the initial model to obtain the prediction percentage. The prediction percentage is used to represent the prompt information corresponding to the training data set and the sample data. After being applied to the initial model, it is obtained by calculating the degree of association between the training data set and the sample data.

In the above step of obtaining a usable prediction model, the server trains the sample data set according to the training parameters to obtain the training data set, and applies the prompt information corresponding to the training data set and the sample data to the initial model to obtain the prediction percentage, and then obtains the prediction percentage based on Sample data sets to build available predictive models. Therefore, by further training the initial model, the available prediction model corresponding to the enterprise is obtained, and the accuracy of risk prediction according to the enterprise index is improved.

It should be understood that although the steps in the flowcharts of FIGS. 2-4 are displayed in order according to the arrows, the steps are not necessarily executed in the order indicated by the arrows. Unless clearly stated in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least some of the steps in FIGS. 2-4 may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but may be executed at different times. These sub-steps or stages The execution order of is not necessarily sequential, but may be executed in turn or alternately with at least a part of other steps or sub-steps or stages of other steps.

In one of the embodiments, as shown in FIG. 5, a risk prediction device is provided, which includes: a dimension information acquisition module 502, an enterprise index extraction module 504, a label acquisition module 506, a prediction information acquisition module 508, and a transmission module 510, where:

The dimension information obtaining module 502 is used to receive the prediction request sent by the terminal and obtain each dimension information of the enterprise corresponding to the prediction request; the prediction request carries the enterprise ID, and the enterprise ID corresponds to each dimension information of the enterprise.

The enterprise index extraction module 504 is used to extract enterprise indexes from all dimensions of the enterprise; the enterprise indexes include various financial indexes, legal litigation information, public opinion information, and import and export lists.

The label acquisition module 506 is used to input the enterprise index as the prediction feature into the available prediction model obtained by pre-training, and output the prediction label corresponding to the enterprise index; the available prediction model is trained based on the prompt information corresponding to the training data set and sample data; prediction Labels include risk prediction labels and risk-free prediction labels.

The prediction information obtaining module 508 is used to obtain the prediction information corresponding to the enterprise index according to the prediction label; the prediction information includes the risk prediction information corresponding to the risk prediction label and the risk-free prediction information corresponding to the risk-free prediction label.

The sending module 510 is used to send the prediction information to the corresponding terminal.

In the above risk prediction device, the server acquires the information of each dimension of the enterprise according to the prediction request sent by the terminal, and extracts the enterprise index from the information of each dimension of the enterprise. Therefore, the enterprise index can be input into the available prediction model that is pre-trained as the prediction feature, the risk prediction information and the risk-free prediction information corresponding to the enterprise index can be obtained, and the risk prediction information and the risk-free prediction information can be sent to the corresponding terminal respectively. In the case of data update, there is no need to repeatedly perform data preprocessing on each indicator, which can reduce resource consumption, and at the same time, send corresponding prediction information to corresponding terminals for different prediction tags, which can further improve the risk prediction effect.

In one of the embodiments, a risk prediction device is provided, and the device further includes:

The prediction label generation module is used to obtain sample data from the database, and extract the sample label from the sample data, and use the sample label as the prediction label of the initial model; the prediction feature generation module is used to obtain the sample index corresponding to the sample data, and The sample index is used as the prediction feature of the initial model; the training data set generation module is used to generate the training data set based on the prediction feature and the prediction label; the available prediction model generation module is used to use the prompt information corresponding to the training data set and the sample data The model is trained to obtain an available prediction model.

In the above risk prediction device, the server extracts the sample label from the acquired sample data, and uses the sample label as the prediction label of the initial model. Obtain the index corresponding to the sample data, and use the index corresponding to the sample data as the prediction feature of the initial model, and generate the training data set according to the prediction feature and the prediction label, and then use the prompt information corresponding to the training data set and the sample data to carry out the initial model Train to get a usable prediction model. Therefore, by directly using the sample label as the prediction label of the initial model, it is possible to reduce the need for reprocessing the sample data every time the data is updated, and reduce resource consumption.

The sample data set acquisition module is used to obtain the sample data set from the database and store the sample data set into a preset object; the training function call module is used to call the training function in the database; the training parameter acquisition module is used to Correspondence between the preset training function and training parameters to obtain the training parameters corresponding to the training function; the initial model building module is used to call the sample data set from the preset object and build according to the training parameters and sample data set The initial model.

In the above risk prediction device, the server obtains the training parameters corresponding to the training function according to the correspondence between the preset training function and the training parameter, and calls the sample data set from the preset object, according to the training parameter and the sample data set, Build the initial model. Therefore, the establishment of the initial model can be achieved, and the training data can be used to train the sample data set, thereby providing a corresponding basis for the subsequent available prediction models, and improving work efficiency.

In one of the embodiments, the dimensional information acquisition module is also used to:

Receive and parse the prediction request sent by the terminal; obtain the enterprise ID carried in the prediction request based on the correspondence between the enterprise ID and each dimension information of the enterprise, and obtain each dimension information of the enterprise corresponding to the enterprise ID.

In the above-mentioned dimension information obtaining module, the server receives and parses the prediction request sent by the terminal, obtains the enterprise ID carried in the prediction request, and obtains each dimension information of the enterprise corresponding to the enterprise ID. Therefore, it is possible to obtain the information of each dimension of the targeted enterprise, clarify the correspondence between different prediction requests and the corresponding enterprise, and facilitate the acquisition and storage of the dimension information of the personalized enterprise.

In one of the embodiments, the prompt information acquisition module is also used to:

Enter enterprise indicators as prediction features into available prediction models; obtain risk prediction tags and risk-free prediction tags corresponding to enterprise indexes; obtain and predict risk tags based on the correspondence between preset risk prediction tags and risk prediction information Corresponding risk prediction information; according to the correspondence between the preset risk-free prediction label and the risk-free prediction information, obtain the risk-free prediction information corresponding to the risk-free prediction label.

In the above prompt information acquisition module, the server inputs enterprise indicators as prediction features into the available prediction model, respectively obtains the risk prediction label and the risk-free prediction label corresponding to the enterprise index, and then can obtain the risk prediction information corresponding to the risk prediction label and Risk-free prediction information corresponding to the risk-free prediction label. Therefore, corresponding prediction labels are obtained for different enterprise indicators, and corresponding risk prediction information and risk-free prediction information can be obtained separately, thereby improving the accuracy of risk prediction.

In one of the embodiments, the predicted label generation module is also used to:

Extract the sample label from the sample data and obtain the attributes of the sample label; divide the sample data into the first type sample and the second type sample according to the attribute; obtain the first sample label corresponding to the first type sample and the second The second sample label corresponding to the class sample; where the first sample label is a sample label carrying risk data, and the second sample label is a sample label without risk data; using the first sample label as the risk prediction label of the initial model, Use the second sample label as the risk-free prediction label of the initial model.

In the above prediction label generation module, the server extracts the sample label from the sample data and obtains the attribute of the sample label, and divides the sample data into the first type sample and the second type sample according to the attribute. Obtain the first sample label corresponding to the first type sample and the second sample label corresponding to the second type sample, and use the first sample label as the risk prediction label of the initial model and the second sample label as the initial model Of risk-free prediction labels. Therefore, the sample tags carrying the risk data and the non-risk data can be distinguished, which is beneficial for performing targeted sample data processing and improving work efficiency.

In one of the embodiments, the prediction information acquisition module is also used to:

According to the corresponding relationship between the preset risk prediction label and the risk prediction information, obtain the risk prediction information corresponding to the risk prediction label; Risk-free prediction information corresponding to the risk prediction label.

In the above prediction information acquisition module, the server can obtain the risk prediction information corresponding to the risk prediction label and the risk-free prediction information corresponding to the risk-free prediction label by separately acquiring the risk prediction label and the risk-free prediction label corresponding to the enterprise index. Therefore, corresponding prediction labels are obtained for different enterprise indicators, and corresponding risk prediction information and risk-free prediction information can be obtained separately, thereby improving the accuracy of risk prediction.

In one of the embodiments, a predictive model building module may be used, which is also used to:

Train the sample data set according to the training parameters to obtain the training data set; apply the prompt information corresponding to the training data set and the sample data to the initial model to obtain the prediction percentage; according to the prediction percentage and the sample data set, construct an available prediction model.

The above-mentioned available prediction model building module, the server trains the sample data set according to the training parameters to obtain the training data set, and applies the prompt information corresponding to the training data set and the sample data to the initial model to obtain the prediction percentage, and then obtains the prediction percentage according to the prediction percentage Data sets, build available predictive models. Therefore, by further training the initial model, the available prediction model corresponding to the enterprise is obtained, and the accuracy of risk prediction according to the enterprise index is improved.

For the specific definition of the risk prediction device, reference may be made to the definition of the risk prediction method above, which will not be repeated here. Each module in the above risk prediction device may be implemented in whole or in part by software, hardware, or a combination thereof. The above modules may be embedded in the hardware or independent of the processor in the computer device, or may be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure may be as shown in FIG. 6. The computer device includes a processor, memory, network interface, and database connected by a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer-readable instructions, and a database. The internal memory provides an environment for the operation of the operating system and computer-readable instructions in the non-volatile storage medium. The database of the computer device is used to store information data of various dimensions of the enterprise. The network interface of the computer device is used to communicate with external terminals through a network connection. The computer readable instructions are executed by the processor to implement a risk prediction method.

Those skilled in the art can understand that the structure shown in FIG. 6 is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer equipment to which the solution of the present application is applied. Include more or less components than shown in the figure, or combine certain components, or have a different arrangement of components.

A computer device includes a memory and one or more processors. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the unbalanced sample data preprocessing method provided in any embodiment of the present application is implemented. A step of.

One or more non-volatile computer-readable storage media storing computer-readable instructions, which when executed by one or more processors, cause the one or more processors to implement any one of the embodiments of the present application The steps of the unbalanced sample data preprocessing method provided.

A person of ordinary skill in the art may understand that all or part of the process in the method of the foregoing embodiments may be completed by instructing relevant hardware through computer-readable instructions, and the computer-readable instructions may be stored in a non-volatile computer In the readable storage medium, when the computer-readable instructions are executed, they may include the processes of the foregoing method embodiments. Wherein, any reference to the memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

The technical features of the above embodiments can be arbitrarily combined. To simplify the description, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, all It is considered as the scope described in this specification.

The above-mentioned embodiments only express several implementations of the present application, and their descriptions are more specific and detailed, but they should not be understood as limiting the scope of the invention patent. It should be noted that, for those of ordinary skill in the art, without departing from the concept of the present application, a number of modifications and improvements can also be made, which all fall within the protection scope of the present application. Therefore, the protection scope of the patent of this application shall be subject to the appended claims.

Claims

A risk prediction method applied to the server, including:

Receiving a prediction request sent by a terminal, and acquiring information of each dimension of the enterprise corresponding to the prediction request; the prediction request carries an enterprise identifier, and the enterprise identifier corresponds to each dimension information of the enterprise;

Extract enterprise indicators from all dimensions of the enterprise; the enterprise indicators include various financial indicators, legal litigation information, public opinion information, and import and export lists;

Input the enterprise index as a prediction feature into an available prediction model obtained by pre-training, and output a prediction label corresponding to the enterprise index; the available prediction model is trained based on the prompt information corresponding to the training data set and sample data; Forecast labels include risk prediction labels and risk-free prediction labels;

Obtaining prediction information corresponding to the enterprise indicator according to the prediction tag; the prediction information includes risk prediction information corresponding to the risk prediction tag, and risk-free prediction information corresponding to the risk-free prediction tag; and

Send the prediction information to the corresponding terminal.
The method according to claim 1, further comprising:

Obtain sample data from the database, and extract sample labels from the sample data, and use the sample labels as the prediction labels of the initial model;

Acquiring a sample index corresponding to the sample data, and using the sample index as a prediction feature of the initial model;

Generating a training data set according to the prediction feature and the prediction label; and

Using the prompt information corresponding to the training data set and the sample data, the initial model is trained to obtain an available prediction model.
The method of claim 2, further comprising:

Obtain the sample data set from the database and store the sample data set in a preset object;

Call the training function in the database;

Obtaining training parameters corresponding to the training function according to the correspondence between the preset training function and the training parameters; and

The sample data set is called from the preset object, and an initial model is constructed according to the training parameters and the sample data set.
The method according to claim 1, wherein the receiving of the prediction request sent by the terminal and obtaining the information of each dimension of the enterprise corresponding to the prediction request includes:

Receive and parse the prediction request sent by the terminal; and

Obtaining the enterprise identification carried in the prediction request;

Based on the correspondence between the enterprise ID and the dimension information of the enterprise, acquire the dimension information of the enterprise corresponding to the enterprise ID.
The method according to claim 2, wherein the extracting the sample label from the sample data and using the sample label as the prediction label of the initial model includes:

Extracting the sample label from the sample data, and obtaining the attributes of the sample label;

Divide the sample data into first-type samples and second-type samples according to the attributes;

Acquiring a first sample label corresponding to the first type of sample and a second sample label corresponding to the second type of sample; wherein the first sample label is a sample label carrying risk data, the The second sample label is a sample label of no risk data; and

The first sample label is used as a risk prediction label of the initial model, and the second sample label is used as a risk-free prediction label of the initial model.
The method according to claim 5, wherein the obtaining prediction information corresponding to the enterprise index according to the prediction tag includes:

Obtain the risk prediction information corresponding to the risk prediction tag according to the correspondence between the preset risk prediction tag and the risk prediction information; and

Obtain the risk-free prediction information corresponding to the risk-free prediction label according to the correspondence between the preset risk-free prediction label and the risk-free prediction information.
The method according to claim 2, wherein the using the prompt information corresponding to the training data set and the sample data to train the initial model to obtain a usable prediction model includes:

Train the sample data set according to the training parameters to obtain the training data set;

Applying the prompt information corresponding to the training data set and the sample data to the initial model to obtain a prediction percentage; and

According to the prediction percentage and the sample data set, the available prediction model is constructed.
A risk prediction device, including:

The dimension information obtaining module is used to receive the prediction request sent by the terminal, and obtain each dimension information of the enterprise corresponding to the prediction request; the prediction request carries an enterprise identifier, and the enterprise identifier corresponds to each dimension information of the enterprise ;

The enterprise index extraction module is used to extract enterprise indexes from all dimensions of the enterprise; the enterprise indexes include various financial indexes, legal litigation information, public opinion information, and import and export lists;

The tag acquisition module is used to input the enterprise index as a prediction feature into an available prediction model that is pre-trained, and output a prediction tag corresponding to the enterprise index; the available prediction model is based on a prompt corresponding to the training data set and sample data Information training; the prediction labels include risk prediction labels and risk-free prediction labels;

The prediction information obtaining module is used to obtain prediction information corresponding to the enterprise index according to the prediction label; the prediction information includes risk prediction information corresponding to the risk prediction label, and corresponding to the risk-free prediction label No risk prediction information; and

The sending module is used to send the prediction information to the corresponding terminal.
The apparatus according to claim 8, wherein the predicted label generation module is further used to:

Extracting the sample label from the sample data, and obtaining the attributes of the sample label;

Divide the sample data into first-type samples and second-type samples according to the attributes;

Acquiring a first sample label corresponding to the first type of sample and a second sample label corresponding to the second type of sample; wherein the first sample label is a sample label carrying risk data, the The second sample label is a sample label of no risk data; and

The first sample label is used as a risk prediction label of the initial model, and the second sample label is used as a risk-free prediction label of the initial model.
A computer device includes a memory and one or more processors. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the one or more processors, the one or more Each processor performs the following steps:

Receiving a prediction request sent by a terminal, and acquiring information of each dimension of the enterprise corresponding to the prediction request; the prediction request carries an enterprise identifier, and the enterprise identifier corresponds to each dimension information of the enterprise;

Extract enterprise indicators from all dimensions of the enterprise; the enterprise indicators include various financial indicators, legal litigation information, public opinion information, and import and export lists;

Input the enterprise index as a prediction feature into an available prediction model obtained by pre-training, and output a prediction label corresponding to the enterprise index; the available prediction model is trained based on the prompt information corresponding to the training data set and sample data; Forecast labels include risk prediction labels and risk-free prediction labels;

Obtaining prediction information corresponding to the enterprise indicator according to the prediction tag; the prediction information includes risk prediction information corresponding to the risk prediction tag, and risk-free prediction information corresponding to the risk-free prediction tag; and

Send the prediction information to the corresponding terminal.
The computer device of claim 10, wherein the processor further executes the following steps when executing the computer-readable instructions:

Obtain sample data from the database, and extract sample labels from the sample data, and use the sample labels as the prediction labels of the initial model;

Acquiring a sample index corresponding to the sample data, and using the sample index as a prediction feature of the initial model;

Generating a training data set according to the prediction feature and the prediction label; and

Using the prompt information corresponding to the training data set and the sample data, the initial model is trained to obtain an available prediction model.
The computer device according to claim 11, wherein the processor further executes the following steps when executing the computer-readable instructions:

Obtain the sample data set from the database and store the sample data set in a preset object;

Call the training function in the database;

Obtaining training parameters corresponding to the training function according to the correspondence between the preset training function and the training parameters; and

The sample data set is called from the preset object, and an initial model is constructed according to the training parameters and the sample data set.
The computer device of claim 10, wherein the processor further executes the following steps when executing the computer-readable instructions:

Receive and parse the prediction request sent by the terminal; and

Obtaining the enterprise identification carried in the prediction request;

Based on the correspondence between the enterprise ID and the dimension information of the enterprise, acquire the dimension information of the enterprise corresponding to the enterprise ID.
The computer device according to claim 11, wherein the processor further executes the following steps when executing the computer-readable instructions:

Extracting the sample label from the sample data, and obtaining the attributes of the sample label;

Divide the sample data into first-type samples and second-type samples according to the attributes;

Acquiring a first sample label corresponding to the first type of sample and a second sample label corresponding to the second type of sample; wherein the first sample label is a sample label carrying risk data, the The second sample label is a sample label of no risk data; and

The first sample label is used as a risk prediction label of the initial model, and the second sample label is used as a risk-free prediction label of the initial model.
The computer device according to claim 14, wherein the processor further executes the following steps when executing the computer-readable instructions:

Obtain the risk prediction information corresponding to the risk prediction tag according to the correspondence between the preset risk prediction tag and the risk prediction information; and

Obtain the risk-free prediction information corresponding to the risk-free prediction label according to the correspondence between the preset risk-free prediction label and the risk-free prediction information.
One or more non-volatile computer-readable storage media storing computer-readable instructions, which when executed by one or more processors, cause the one or more processors to perform the following steps:

Receiving a prediction request sent by a terminal, and acquiring information of each dimension of the enterprise corresponding to the prediction request; the prediction request carries an enterprise identifier, and the enterprise identifier corresponds to each dimension information of the enterprise;

Extract enterprise indicators from all dimensions of the enterprise; the enterprise indicators include various financial indicators, legal litigation information, public opinion information, and import and export lists;

Input the enterprise index as a prediction feature into an available prediction model obtained by pre-training, and output a prediction label corresponding to the enterprise index; the available prediction model is trained based on the prompt information corresponding to the training data set and sample data; Forecast labels include risk prediction labels and risk-free prediction labels;

Obtaining prediction information corresponding to the enterprise indicator according to the prediction tag; the prediction information includes risk prediction information corresponding to the risk prediction tag, and risk-free prediction information corresponding to the risk-free prediction tag; and

Send the prediction information to the corresponding terminal.
The storage medium according to claim 16, wherein when the computer-readable instructions are executed by the processor, the following steps are further performed:

Obtain sample data from the database, and extract sample labels from the sample data, and use the sample labels as the prediction labels of the initial model;

Acquiring a sample index corresponding to the sample data, and using the sample index as a prediction feature of the initial model;

Generating a training data set according to the prediction feature and the prediction label; and

Using the prompt information corresponding to the training data set and the sample data, the initial model is trained to obtain an available prediction model.
The storage medium according to claim 17, wherein when the computer-readable instructions are executed by the processor, the following steps are further performed:

Obtain the sample data set from the database and store the sample data set in a preset object;

Call the training function in the database;

Obtaining training parameters corresponding to the training function according to the correspondence between the preset training function and the training parameters; and

The sample data set is called from the preset object, and an initial model is constructed according to the training parameters and the sample data set.
The storage medium according to claim 16, wherein when the computer-readable instructions are executed by the processor, the following steps are further performed:

Receive and parse the prediction request sent by the terminal; and

Obtaining the enterprise identification carried in the prediction request;

Based on the correspondence between the enterprise ID and the dimension information of the enterprise, acquire the dimension information of the enterprise corresponding to the enterprise ID.
The storage medium according to claim 17, wherein when the computer-readable instructions are executed by the processor, the following steps are further performed:

Extracting the sample label from the sample data, and obtaining the attributes of the sample label;

Divide the sample data into first-type samples and second-type samples according to the attributes;

Acquiring a first sample label corresponding to the first type of sample and a second sample label corresponding to the second type of sample; wherein the first sample label is a sample label carrying risk data, the The second sample label is a sample label of no risk data; and

The first sample label is used as a risk prediction label of the initial model, and the second sample label is used as a risk-free prediction label of the initial model.