CN114612251A

CN114612251A - Risk assessment method, device, equipment and storage medium

Info

Publication number: CN114612251A
Application number: CN202210246481.9A
Authority: CN
Inventors: 夏龙江
Original assignee: Ping An Technology Shenzhen Co Ltd
Current assignee: Ping An Technology Shenzhen Co Ltd
Priority date: 2022-03-14
Filing date: 2022-03-14
Publication date: 2022-06-10

Abstract

The invention relates to the technical field of big data, and discloses a risk assessment method, a risk assessment device, risk assessment equipment and a storage medium. The method comprises the following steps: acquiring original service data, and cleaning the service data to obtain standard service data; acquiring historical service data according to the data type of the standard service data, and configuring a wind control rule according to the historical service data; performing dimension splitting on the service data according to a preset dimension to generate a plurality of target service data; determining a wind control evaluation factor corresponding to the target service data according to a wind control rule; and inputting the target service data and the risk assessment factor into a preset wind control model for analysis to obtain a risk assessment result of the service data. According to the invention, the risk in the reimbursement process is intercepted through the dynamic configuration of reimbursement rules, and the risk interception result is verified, so that the user loss is reduced. The wind control efficiency is improved.

Description

Risk assessment method, device, equipment and storage medium

Technical Field

The invention relates to the technical field of big data, in particular to a risk assessment method, a risk assessment device, risk assessment equipment and a storage medium.

Background

In various industries, fields and channels, it is not sufficient to ensure that things develop in a good direction and to avoid unpredictable economic and property losses. At this time, a complete set of wind control system is developed to solve various problems in actual production business. As the subject of things, various measures and methods can be taken to eliminate or reduce various possibilities of occurrence of risk events, or to reduce losses caused when risk events occur.

At present, the existing reimbursement risk management and control system has fixed rules, is difficult to expand, cannot achieve dynamic updating, mostly does not perform advanced and posterior dual management and control, is insufficient in safety degree, solves the pain points, and improves user experience.

Disclosure of Invention

The invention mainly aims to provide a risk assessment method, a risk assessment device, risk assessment equipment and a storage medium. The wind control efficiency is improved.

The first aspect of the present invention provides a risk assessment method, including: acquiring original service data, and cleaning the service data to obtain standard service data; acquiring historical service data according to the data type of the standard service data, and configuring a wind control rule according to the historical service data; performing dimension splitting on the business data according to a preset dimension to generate a plurality of target business data; determining a wind control evaluation factor corresponding to the target service data according to the wind control rule; and inputting the target service data and the risk assessment factor into a preset wind control model for analysis to obtain a risk assessment result of the service data.

Optionally, in a first implementation manner of the first aspect of the present invention, the obtaining original service data and cleaning the service data to obtain standard service data includes: carrying out duplication removal operation on the original service data, and detecting whether the original service data after duplication removal has a data missing value or not; if the missing data value does not exist, the original service data after the duplication removal is used as standard service data; and if the data missing value exists, performing data filling on the data missing value to obtain standard service data.

Optionally, in a second implementation manner of the first aspect of the present invention, the obtaining historical service data according to a data type of the standard service data, and configuring a wind control rule according to the historical service data includes: determining the type of the service data, acquiring historical service data based on the type, extracting a wind control analysis result in the historical service data, and determining a risk threshold; adjusting a preset wind control strategy according to the risk threshold value to obtain an actual wind control strategy; determining a wind control effect representation value corresponding to the actual wind control strategy according to the actual wind control strategy and the wind control analysis result; and configuring a wind control rule according to the wind control effect representation value and the historical service data.

Optionally, in a third implementation manner of the first aspect of the present invention, the configuring a wind control rule according to the wind control policy corresponding to the target wind control effect representation value and the historical service data includes: determining a target wind control effect representation value corresponding to the wind control strategy according to the wind control strategy and the historical service data; and taking the target wind control effect representation value as a test parameter, and configuring a wind control rule according to the target range and the historical service data.

Optionally, in a fourth implementation manner of the first aspect of the present invention, the performing dimension splitting on the service data according to a preset dimension, and generating multiple target service data includes: acquiring a data field of the standard service data, and identifying the data attribute of the standard service data according to the data field; clustering the standard service data with the same data attribute to obtain a data clustering center point; and generating corresponding target service data according to the data clustering center point.

Optionally, in a fifth implementation manner of the first aspect of the present invention, before the inputting the target service data and the risk assessment factor into a preset wind control model for analysis to obtain a risk assessment result of the service data, the method further includes: acquiring sample characteristic data of a plurality of wind control sample users, wind control performance data of the wind control sample users in a scene to be wind controlled and wind controlled duration information of the wind control sample users in the scene to be wind controlled; determining a wind control evaluation label of the wind control sample user according to the wind control performance data; processing the sample characteristic data according to the wind control evaluation label to obtain target characteristic data of the wind control sample user; dividing the target characteristic data, the wind-controlled duration information and the wind-controlled evaluation label according to a preset proportion to obtain training sample data and test sample data; inputting the training sample data into a preset convolutional neural network model for training to obtain an original wind control model; and inputting the test sample data into the original wind control model for testing to obtain a target wind control model.

Optionally, in a sixth implementation manner of the first aspect of the present invention, inputting the training sample data into a preset convolutional neural network model for training, and obtaining an original wind control model includes: initializing model parameters of a convolutional neural network model; performing feature extraction on the training sample data by adopting a convolutional neural network to obtain a feature vector of the training sample data; and inputting the feature vector of the training sample data into the convolutional neural network model for training to obtain the original wind control model.

A second aspect of the present invention provides a risk assessment apparatus, comprising: the data cleaning module is used for acquiring original service data and cleaning the service data to obtain standard service data; the configuration module is used for acquiring historical service data according to the data type of the standard service data and configuring a wind control rule according to the historical service data; the generating module is used for carrying out dimension splitting on the service data according to a preset dimension to generate a plurality of target service data; the first determining module is used for determining a wind control evaluation factor corresponding to the target service data according to the wind control rule; and the analysis module is used for inputting the target service data and the risk assessment factor into a preset wind control model for analysis to obtain a risk assessment result of the service data.

Optionally, in a first implementation manner of the second aspect of the present invention, the data cleansing module is specifically configured to: carrying out duplication removal operation on the original service data, and detecting whether the original service data after duplication removal has a data missing value or not; if the missing data value does not exist, the original service data after the duplication removal is used as standard service data; and if the data missing value exists, performing data filling on the data missing value to obtain standard service data.

Optionally, in a second implementation manner of the second aspect of the present invention, the configuration module includes: the extraction unit is used for determining the type of the service data, acquiring historical service data based on the type, extracting a wind control analysis result in the historical service data and determining a risk threshold; the adjusting unit is used for adjusting a preset wind control strategy according to the risk threshold value to obtain an actual wind control strategy; the determining unit is used for determining a wind control effect representation value corresponding to the actual wind control strategy according to the actual wind control strategy and the wind control analysis result; and the configuration unit is used for configuring a wind control rule according to the wind control effect representation value and the historical service data.

Optionally, in a third implementation manner of the second aspect of the present invention, the configuration unit is specifically configured to:

determining a wind control effect representation value corresponding to the actual wind control strategy according to the actual wind control strategy and the historical service data; and taking the wind control effect characteristic value as a test parameter, and configuring a wind control rule according to a target range corresponding to a preset wind control effect characteristic value and the historical service data.

Optionally, in a fourth implementation manner of the second aspect of the present invention, the generating module is specifically configured to: acquiring a data field of the standard service data, and identifying the data attribute of the standard service data according to the data field; clustering the standard service data with the same data attribute to obtain a data clustering center point; and generating corresponding target service data according to the data clustering central point.

Optionally, in a fifth implementation manner of the second aspect of the present invention, the risk assessment apparatus further includes: the system comprises an acquisition module, a judgment module and a display module, wherein the acquisition module is used for acquiring sample characteristic data of a plurality of wind control sample users, wind control performance data of the wind control sample users in a scene to be wind controlled and wind controlled duration information of the wind control sample users in the scene to be wind controlled; the second determination module is used for determining a wind control evaluation label of the wind control sample user according to the wind control performance data; processing the sample characteristic data according to the wind control evaluation label to obtain target characteristic data of the wind control sample user; the dividing module is used for dividing the target characteristic data, the controlled duration information and the wind control evaluation label according to a preset proportion to obtain training sample data and test sample data; the training module is used for inputting the training sample data into a preset convolutional neural network model for training to obtain an original wind control model; and the test module is used for inputting the test sample data into the original wind control model for testing to obtain a target wind control model.

Optionally, in a sixth implementation manner of the second aspect of the present invention, the training module is specifically configured to: initializing model parameters of a convolutional neural network model; performing feature extraction on the training sample data by adopting a convolutional neural network to obtain a feature vector of the training sample data; and inputting the feature vector of the training sample data into the convolutional neural network model for training to obtain the original wind control model.

A third aspect of the present invention provides a risk assessment apparatus comprising: a memory having instructions stored therein and at least one processor, the memory and the at least one processor interconnected by a line;

the at least one processor invokes the instructions in the memory to cause the risk assessment device to perform the steps of the risk assessment method described above.

A fourth aspect of the present invention provides a computer-readable storage medium having stored therein instructions, which, when run on a computer, cause the computer to perform the steps of the above-described risk assessment method.

In the technical scheme provided by the invention, standard service data are obtained by acquiring original service data and cleaning the service data; acquiring historical service data according to the data type of the standard service data, and configuring a wind control rule according to the historical service data; performing dimension splitting on the service data according to a preset dimension to generate a plurality of target service data; determining a wind control evaluation factor corresponding to the target service data according to a wind control rule; and inputting the target service data and the risk assessment factor into a preset wind control model for analysis to obtain a risk assessment result of the service data. According to the invention, the risk in the reimbursement process is intercepted through the dynamic configuration of reimbursement rules, and the risk interception result is verified, so that the user loss is reduced. The wind control efficiency is improved.

Drawings

FIG. 1 is a schematic diagram of a risk assessment method according to a first embodiment of the present invention;

FIG. 2 is a schematic diagram of a risk assessment method according to a second embodiment of the present invention;

FIG. 3 is a schematic diagram of a risk assessment method provided by the present invention;

FIG. 4 is a schematic diagram of a fourth embodiment of the risk assessment method provided by the present invention;

FIG. 5 is a schematic diagram of a fifth embodiment of the risk assessment method provided by the present invention;

FIG. 6 is a schematic view of a risk assessment device according to a first embodiment of the present invention;

FIG. 7 is a schematic view of a risk assessment device according to a second embodiment of the present invention;

fig. 8 is a schematic diagram of an embodiment of a risk assessment device provided by the present invention.

Detailed Description

According to the risk assessment method, the risk assessment device, the risk assessment equipment and the risk assessment storage medium, the original business data are obtained, and the business data are cleaned to obtain the standard business data; acquiring historical service data according to the data type of the standard service data, and configuring a wind control rule according to the historical service data; performing dimension splitting on the service data according to a preset dimension to generate a plurality of target service data; determining a wind control evaluation factor corresponding to the target service data according to a wind control rule; and inputting the target service data and the risk assessment factor into a preset wind control model for analysis to obtain a risk assessment result of the service data. According to the invention, the risk in the reimbursement process is intercepted through the dynamic configuration of reimbursement rules, and the risk interception result is verified, so that the user loss is reduced. The wind control efficiency is improved.

The terms "first," "second," "third," "fourth," and the like in the description and in the claims, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It will be appreciated that the data so used may be interchanged under appropriate circumstances such that the embodiments described herein may be practiced otherwise than as specifically illustrated or described herein. Furthermore, the terms "comprises," "comprising," or "having," and any variations thereof, are intended to cover non-exclusive inclusions, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

For understanding, the following describes a specific process of an embodiment of the present invention, and with reference to fig. 1, a first embodiment of a risk assessment method according to an embodiment of the present invention includes:

101. acquiring original service data, and cleaning the service data to obtain standard service data;

in this embodiment, the original service data is generated based on different service scenarios, for example, if the service scenario is an enterprise loan approval scenario, the original service data includes: enterprise data including business data, judicial data, public opinion data, upstream customer data, downstream customer data, financial data, stream data, industry data, etc., and enterprise shareholder data including personal credit investigation data, loan data, asset data, etc. Further, in this embodiment, the original service data may be acquired offline by service personnel; the wind control data can be collected on line through a collection tool, such as a crawler tool, so that the diversity of wind control data sources is realized, the comprehensiveness of the wind control data is guaranteed, and the wind control analysis accuracy of subsequent data is guaranteed.

Further, it should be understood that some useless data and repeated data exist in the collected original service data, and in order to improve the rapidity of generating a subsequent data wind control analysis report, the embodiment of the invention performs data cleaning on the original service data so as to reduce the data volume of subsequent data analysis.

Specifically, in an optional embodiment of the present invention, the performing data cleaning on the original service data to obtain standard service data includes: carrying out duplication removal operation on the original service data, and detecting whether the original service data after duplication removal has a data missing value or not; if the missing data value does not exist, the original service data after the duplication removal is used as standard service data; and if the data missing value exists, performing data filling on the data missing value to obtain standard service data.

Further, the performing a deduplication operation on the original service data includes: and calculating the similarity of any two pieces of wind control data in the original service data, if the similarity is not greater than the preset similarity, simultaneously retaining the two pieces of wind control data, and if the similarity is greater than the preset similarity, deleting any one of the two pieces of wind control data.

102. Acquiring historical service data according to the data type of the standard service data, and configuring a wind control rule according to the historical service data;

in this embodiment, when the wind control personnel need to configure the wind control policy according to the actual wind control requirement, a target is generally set, that is, what wind control effect can be achieved by the configured wind control policy. Therefore, in the embodiment of the present application, a target range of the wind control effect representation value may be set according to an actual wind control requirement, and the target range is input into the server.

Specifically, after acquiring each historical service data, the server needs to determine, for each acquired historical service data, a service data amount corresponding to the historical service data, further determine a total service data amount corresponding to each historical service data, and further determine each risk threshold according to the determined total service data amount. The traffic data amount corresponding to the historical traffic data may be included in the historical traffic data.

Further, after the risk threshold values are determined, a plurality of wind control strategies can be determined according to preset risk probabilities. The preset risk probabilities mentioned here may be all the risk probabilities that the wind control model can output. In practical applications, although the risk probabilities output by different wind control models are different in probability standard, the risk probabilities that can be output by the wind control models may be the same in value. After each wind control strategy is determined, the server can respectively determine the wind control effect representation value corresponding to each wind control strategy. After determining each wind control effect characteristic value corresponding to each wind control strategy, the server can select the wind control effect characteristic value falling into the target range of the wind control effect characteristic value, and configures the wind control strategy corresponding to the wind control effect characteristic value in the subsequent process. And when the wind control effect representation value comprises the coverage rate and the disturbance rate, the server should select a wind control strategy of which the coverage rate and the disturbance rate both fall within the corresponding target range. .

103. Performing dimension splitting on the service data according to a preset dimension to generate a plurality of target service data;

in this embodiment, because the data in the standard service data may have different attributes, the standard service data may have data with different dimensions, that is, data of different types, such as time dimension data, basic information dimension data, and behavior dimension data, so that the embodiment of the present invention performs dimension splitting on the standard service data to merge the data with the same attribute into one type, which is convenient for subsequent data calculation and improves the efficiency of data analysis.

Specifically, in another optional embodiment, the performing dimension splitting on the standard business data to generate a plurality of target business data includes: and acquiring a data field of the standard service data, identifying the data attribute of the standard service data according to the data field, clustering the standard service data with the same data attribute to obtain a data clustering central point, and generating corresponding target service data according to the data clustering central point. The data field is an entity object parameter used for representing the standard service data, and the data attribute is used for representing the data type of the standard service data, such as a data name, a data feature, and the like.

104. Determining a wind control evaluation factor corresponding to the target service data according to a wind control rule;

in this embodiment, the wind control evaluation factor refers to an analysis rule corresponding to the wind control data, and includes an analysis logic and an analysis mode of the wind control data. The wind control Template comprises a Freemarker Language (FTL) component, namely a bottom Template engine of wind control data, and is used for generating output texts such as document texts, html texts, mail texts and the like from data loaded into the wind control Template.

In another embodiment, the constructing a wind control evaluation factor for each of the target traffic data includes: acquiring the data attribute of the target service data, configuring the data analysis logic of the target service data according to the data attribute, determining the data analysis mode of the target service data, and generating the wind control evaluation factor of the target service data according to the data analysis logic and the data analysis mode.

105. And inputting the target service data and the risk assessment factor into a preset wind control model for analysis to obtain a risk assessment result of the service data.

In this embodiment, a forward propagation algorithm is used to train the feature vector to obtain the first state parameter. Specifically, the training of the face features by using a Forward Propagation (Forward Propagation) algorithm refers to training according to the sequence of time sequence states carried by the face features by using a Forward Propagation algorithm. The first state parameter refers to a parameter obtained in an initial iteration process of model training based on the human face features.

The Forward Propagation (Forward Propagation) algorithm is an algorithm for training a model according to a time sequence. Specifically, the calculation formula of the forward propagation algorithm is sum, wherein St represents the output of the hidden layer at the current moment; representing the weight from the previous moment to the current moment of the hidden layer; representing the weight from the input layer to the hidden layer; a prediction output representing a current time; representing the weights of the hidden layer to the output layer.

In the embodiment of the invention, standard service data are obtained by acquiring original service data and cleaning the service data; acquiring historical service data according to the data type of the standard service data, and configuring a wind control rule according to the historical service data; performing dimension splitting on the service data according to a preset dimension to generate a plurality of target service data; determining a wind control evaluation factor corresponding to the target service data according to a wind control rule; and inputting the target service data and the risk assessment factor into a preset wind control model for analysis to obtain a risk assessment result of the service data. According to the invention, the risk in the reimbursement process is intercepted through the dynamic configuration of reimbursement rules, and the risk interception result is verified, so that the user loss is reduced. The wind control efficiency is improved.

Referring to fig. 2, a second embodiment of the risk assessment method according to the embodiment of the present invention includes:

201. acquiring original service data, and cleaning the service data to obtain standard service data;

202. determining the type of the service data, acquiring historical service data based on the type, extracting a wind control analysis result in the historical service data, and determining a risk threshold;

in this embodiment, in practical application, when a wind control worker needs to configure a wind control policy according to an actual wind control requirement, a target is usually set, that is, what wind control effect can be achieved by the configured wind control policy. Therefore, in the embodiment of the present application, a target range of the wind control effect representation value may be set according to an actual wind control requirement, and the target range is input into the server.

The wind control effect characteristic value mentioned here is a numerical value capable of quantifying the wind control effect, and the wind control effect characteristic value can reflect the wind control effect that the wind control strategy can achieve. For example, the wind control effect characterization values may be a coverage rate and a disturbance rate. The coverage rate represents the ratio of the risk-existing service information which can be identified by the wind control strategy to all the actual risk-existing service information in the actual risk-existing service information, so that the coverage rate can effectively reflect the risk identification capability of the wind control strategy as a wind control effect representation value, and the higher the coverage rate is, the stronger the risk identification capability of the wind control strategy is. The disturbance rate is expressed as the ratio of the number of the business information with risks to the total number of all the business information, wherein the actual wind control result is safe, and the server identifies the number of the business information with risks through a wind control strategy. The higher the disturbance rate is, the higher the possibility of false identification when the server identifies risks through a wind control strategy is.

The target range of the wind control effect representation value mentioned here refers to the wind control effect that the wind control strategy is expected to achieve. The target range of the wind control effect representation value may be a value range or a numerical value, for example, the target range of the coverage rate is 92%.

In this embodiment, there may be many ways for the server to obtain the target range, for example, the server may obtain the target range by inputting the target range of the wind control effect representation value by the wind control staff, or obtain the target range by the option selected by the wind control staff, which is not illustrated herein.

In this embodiment, the server needs to obtain historical service data in addition to the target range of the wind control effect representation value, so as to determine a corresponding wind control strategy through the obtained historical service data and the actual wind control result thereof in the subsequent process.

After the server acquires each historical service data, it needs to determine, for each acquired historical service data, a service data amount corresponding to the historical service data, further determine a total service data amount corresponding to each historical service data, and further determine each risk threshold according to the determined total service data amount. The traffic data amount corresponding to the historical traffic data may be included in the historical traffic data.

For example, still taking the transaction service as an example, it is assumed that the historical service data acquired by the server is a historical transaction request initiated by each user to the server within a past period of time. After the server obtains the historical transaction requests, the server can determine the transaction amount related in the historical transaction requests according to each historical transaction request, wherein the transaction amount is the service data volume corresponding to the historical service data. After the server determines the transaction amount related to each historical transaction request, the server may determine the sum of the transaction amounts (i.e., the total transaction amount) of all the obtained historical transaction requests, where the total transaction amount is the total amount of the service data.

203. Adjusting a preset wind control strategy according to the risk threshold value to obtain an actual wind control strategy;

in this embodiment, after each risk threshold is determined, a plurality of wind control strategies may be determined according to each preset risk probability. The preset risk probabilities mentioned here may be all the risk probabilities that the wind control model can output. In practical applications, although the risk probabilities output by different wind control models are different in probability standard, the risk probabilities that can be output by the wind control models may be the same in value. For example, for different wind control models, the risk probabilities that the wind control models can output are all 0.01-1, and the difference between every two adjacent risk probabilities is 0.01.

Each wind control strategy may include a risk probability and a risk threshold corresponding thereto, and different wind control strategies are obtained by combining different risk probabilities and risk thresholds. For example, assuming that the server determines 100 risk thresholds of 100 to 10000 (the difference between every two adjacent risk thresholds is 100), the preset risk probabilities are 0.01 to 1 (the difference between every two adjacent risk probabilities is 0.01). The server can combine the 100 risk thresholds and the 100 risk probabilities pairwise to obtain 10000 combinations of risk probabilities and risk thresholds, wherein each combination is a wind control strategy.

204. Determining a wind control effect representation value corresponding to the actual wind control strategy according to the actual wind control strategy and the wind control analysis result;

in this embodiment, after each wind control policy is determined, the server may determine the wind control effect representation value corresponding to each wind control policy respectively. The specific determination method may be: for each wind control strategy, the server may determine, according to the obtained traffic data volume and normalized risk probability corresponding to each historical traffic data, and the risk threshold and risk probability included in the wind control strategy, historical traffic data that satisfy a first set condition from each historical traffic data, and determine, according to the determined historical traffic data and the actual wind control result corresponding to the historical traffic data, a wind control effect representation value corresponding to the wind control strategy.

In this embodiment, the wind control effect characterizing value may include an interference rate and a coverage rate, where when the wind control effect characterizing value is the interference rate, the server may determine, according to the traffic data amount and the normalized risk probability corresponding to each historical traffic data, historical traffic data whose normalized risk probability is greater than the risk probability included in the wind control policy and whose traffic data amount is greater than the risk threshold included in the wind control policy from among the obtained historical traffic data.

205. Configuring a wind control rule according to the wind control effect representation value and historical service data;

in this embodiment, after determining each wind control effect characteristic value corresponding to each wind control strategy, the server may select a wind control effect characteristic value falling within a target range of the wind control effect characteristic value, and configure the wind control strategy corresponding to the wind control effect characteristic value in a subsequent process. And when the wind control effect representation value comprises the coverage rate and the disturbance rate, the server should select a wind control strategy of which the coverage rate and the disturbance rate both fall within the corresponding target range. For example, assume that the target range of the wind control effect characterization value is: the coverage rate is 85% -100%, the disturbance rate is 0-2%, and the wind control strategy that the coverage rate is 85% -100% and the disturbance rate is 0-2% needs to be selected and configured by the server.

When the server determines that the wind control effect characteristic value falling into the target range does not exist through the method, the server can determine the wind control score corresponding to each wind control strategy according to the wind control effect characteristic value corresponding to each determined wind control strategy, the target range of the wind control effect characteristic value and a preset wind control scoring mode, select the wind control score meeting a second set condition from the wind control scores, and then configure the wind control strategy corresponding to the wind control score in the subsequent process.

In this embodiment, if the server determines the wind control effect characteristic value corresponding to each wind control policy directly according to the acquired historical service data of all the users and the actual wind control results thereof, the wind control policy corresponding to the selected wind control effect characteristic value is also applicable to all the users. Therefore, the selected wind control strategy can be directly configured, and then in the subsequent process, the wind control strategy is used for wind control on the service requests sent by all the users.

If the server selects a user group from all users as a first user group and selects a wind control strategy for the first user group according to the first user group, the wind control strategy cannot be directly configured for all users, but wind control test needs to be performed on each acquired historical service data through the wind control strategy so as to determine whether the wind control strategy is suitable for wind control of all users.

In this embodiment, after the wind control policy for the first user group is selected, the server may determine, for each of the other user groups except the first user group, a wind control effect characterization value of the wind control policy for the other user group and use the wind control effect characterization value as a test parameter. And when the test parameter is determined to fall into the target range of the wind control effect representation value, configuring the wind control strategy so as to be used for wind control on the users with the same category as the other user groups.

And when the server determines that the test parameters do not fall into the target range of the wind control effect representation value, configuring the wind control strategy selected for the first user group so as to be used for wind control on users with the same category as the first user group in the subsequent process. Meanwhile, the server can determine a historical wind control strategy corresponding to each other user group according to the acquired historical service data. The historical wind control strategy is a wind control strategy used when wind control is performed on the users of the other user groups.

The server can adjust the historical wind control strategy in a preset adjusting mode, so that the adjusted historical wind control strategy obtains a wind control effect characteristic value for the other user groups and falls into a target range of the wind control effect characteristic value. In other words, the server may perform a wind control test on the historical service data corresponding to the other user groups through the adjusted historical wind control policy, and the obtained wind control effect characteristic value may fall within a target range of the wind control effect characteristic value.

206. Performing dimension splitting on the service data according to a preset dimension to generate a plurality of target service data;

207. determining a wind control evaluation factor corresponding to the target service data according to a wind control rule;

208. and inputting the target service data and the risk assessment factors into a preset wind control model for analysis to obtain a risk assessment result of the service data.

The steps 207-210 in the present embodiment are similar to the steps 102-105 in the first embodiment, and are not described herein again.

In the embodiment of the invention, standard service data are obtained by acquiring original service data and cleaning the service data; acquiring historical service data according to the data type of the standard service data, and configuring a wind control rule according to the historical service data; performing dimension splitting on the service data according to a preset dimension to generate a plurality of target service data; determining a wind control evaluation factor corresponding to the target service data according to a wind control rule; and inputting the target service data and the risk assessment factors into a preset wind control model for analysis to obtain a risk assessment result of the service data. According to the invention, the risk in the reimbursement process is intercepted through the dynamic configuration of reimbursement rules, and the risk interception result is verified, so that the user loss is reduced. The wind control efficiency is improved.

Referring to fig. 3, a third embodiment of the risk assessment method according to the embodiment of the present invention includes:

301. acquiring original service data, and cleaning the service data to obtain standard service data;

302. carrying out duplicate removal operation on the original service data, and detecting whether the original service data subjected to duplicate removal has a data missing value or not;

in this embodiment, the original service data is generated based on different service scenarios, for example, if the service scenario is an enterprise loan approval scenario, the original service data includes: enterprise data including business data, judicial data, public opinion data, upstream customer data, downstream customer data, financial data, stream data, industry data, etc., and enterprise shareholder data including personal credit investigation data, loan data, asset data, etc. Further, in an optional embodiment of the present invention, the original service data is collected through the following three ways: the method comprises the following steps of firstly, querying a background database of a professional webpage, such as a sky-eye query; the second mode is offline collection of service personnel; and the third mode is to collect the tools on line, such as a crawler tool, so as to realize the diversity of wind control data sources, ensure the comprehensiveness of the wind control data and further ensure the wind control analysis accuracy of subsequent data.

303. If the missing data value does not exist, the original service data after the duplication removal is used as standard service data;

in the embodiment, some useless data and repeated data exist in the acquired original service data, and in order to improve the rapidity of generating the subsequent data wind control analysis report, the embodiment of the invention performs data cleaning on the original service data so as to reduce the data volume of the subsequent data analysis.

Specifically, in another optional embodiment, the performing data cleaning on the original service data to obtain standard service data includes: carrying out duplication removal operation on the original service data, and detecting whether the original service data after duplication removal has a data missing value or not; if the missing data value does not exist, the original service data after the duplication removal is used as standard service data; and if the data missing value exists, performing data filling on the data missing value to obtain standard service data.

304. If the data missing value exists, performing data filling on the data missing value to obtain standard service data;

in this embodiment, the performing the deduplication operation on the original service data includes: and calculating the similarity of any two pieces of wind control data in the original service data, if the similarity is not greater than the preset similarity, simultaneously retaining the two pieces of wind control data, and if the similarity is greater than the preset similarity, deleting any one of the two pieces of wind control data.

It should be noted that, before calculating the similarity of the original service data, the embodiment of the present invention further includes: and converting the original service data into a corresponding hash value by using a hash algorithm so as to realize the calculation of the similarity of the subsequent original service data.

305. Acquiring a data field of standard service data, and identifying the data attribute of the standard service data according to the data field;

306. Clustering the standard service data with the same data attribute to obtain a data clustering center point;

in this embodiment, the performing the dimension splitting on the standard service data to generate a plurality of target service data includes: and acquiring a data field of the standard service data, identifying the data attribute of the standard service data according to the data field, clustering the standard service data with the same data attribute to obtain a data clustering central point, and generating corresponding target service data according to the data clustering central point.

307. Generating corresponding target service data according to the data clustering center point;

in this embodiment, the data field is an entity object parameter used for characterizing the standard service data, and the data attribute is used for characterizing a data type of the standard service data, such as a data name and a data feature.

308. Determining a wind control evaluation factor corresponding to the target service data according to a wind control rule;

309. and inputting the target service data and the risk assessment factor into a preset wind control model for analysis to obtain a risk assessment result of the service data.

Steps 301 and 308-309 in this embodiment are similar to steps 101 and 104-105 in the first embodiment, and are not described herein again.

Referring to fig. 4, a fourth embodiment of the risk assessment method according to the embodiment of the present invention includes:

401. acquiring original service data, and cleaning the service data to obtain standard service data;

402. acquiring historical service data according to the data type of the standard service data, and configuring a wind control rule according to the historical service data;

403. performing dimension splitting on the service data according to a preset dimension to generate a plurality of target service data;

404. determining a wind control evaluation factor corresponding to the target service data according to a wind control rule;

405. acquiring sample characteristic data of a plurality of wind control sample users, wind control expression data of the wind control sample users in a scene to be wind controlled and wind controlled duration information of the wind control sample users in the scene to be wind controlled;

in this embodiment, the reimbursement duration of each candidate reimbursement user may be obtained, and each candidate reimbursement user may be screened according to the reimbursement duration of each candidate reimbursement user of the preset duration threshold, so as to obtain the sample reimbursement user. The preset time threshold may be one month or more than one month, may also be a preset number of days, and may also be a number of reimbursement periods, which is not particularly limited in the present invention.

Taking a preset time threshold as one month as an example, obtaining reimbursement time lengths of a plurality of candidate reimbursement users, for example, the reimbursement time length of the candidate reimbursement user 1 is 0 month, the reimbursement time length of the candidate reimbursement user 2 is 3 months, and the reimbursement time length of the candidate reimbursement user 3 is 1 month. Taking the candidate reimbursement users with reimbursement duration greater than or equal to the preset duration threshold as sample reimbursement users, for example, screening the three candidate reimbursement users, and taking the candidate reimbursement user 2 and the candidate reimbursement user 3 as sample reimbursement users.

Here, the reimbursement duration of 0 month may be that the number of days from the loan date to the sample collection date of the candidate reimbursement user 1 is zero, that is, the candidate reimbursement user 1 successfully applies for the loan on the sample collection date; a reimbursement duration of 0 months may also refer to a number of days from the deposit day to the sample collection day of less than one month. The reimbursement duration of each of the other candidate reimbursement users is similar to the reimbursement duration of the candidate reimbursement user 1, and is not described herein again.

406. Determining a wind control evaluation label of a wind control sample user according to the wind control performance data;

in this embodiment, the wind control evaluation tag may be used to characterize whether a user of the wind control sample has a risk. In the reimbursement wind control scene, the wind control evaluation label can be an overdue severity label, the overdue severity label comprises a first label and a second label, the first label is used for representing that the sample feature user is seriously overdue, and the second label is used for representing that the sample feature user is not seriously overdue.

Optionally, determining a wind control evaluation tag of a wind control sample user according to the wind control performance data includes: extracting risk behavior data of a wind control sample user from the wind control performance data; the risk behavior data comprises risk behavior duration and/or risk behavior times; performing rolling rate analysis on the risk behavior data to obtain a risk behavior data threshold; and determining a wind control evaluation label of the wind control sample user according to the risk behavior data and the risk behavior data threshold value of the wind control sample user.

In one embodiment, for each wind control sample user, the wind control sample user is bound with a corresponding wind control duration label and a corresponding wind control evaluation label.

407. Processing the sample characteristic data according to the wind control evaluation label to obtain target characteristic data of a wind control sample user;

in this embodiment, according to the wind control evaluation label of the wind control sample user, chi-square binning, screening and encoding are performed on the sample characteristic data of the wind control sample user to obtain target characteristic data of the wind control sample user.

When the variable values of the same sample characteristic variable of a plurality of sample characteristic users are close, the influence of the variable values on the wind control model is very little, so that the target wind control sample users with the variable values belonging to a certain variable value range can be determined, and the variable values of all the target wind control sample users are converted into the code values corresponding to the variable value range. The target characteristic data is obtained by performing chi-square binning, screening and encoding on the sample characteristic data, and modeling is performed by using the target characteristic data, so that the complexity of a wind control model can be simplified, and the calculated amount is reduced.

Optionally, the sample characteristic data includes variable values of a plurality of sample characteristic variables; according to the wind control evaluation label of the wind control sample user, processing the sample characteristic data of the wind control sample user to obtain the target characteristic data of the wind control sample user, and the method comprises the following steps: performing chi-square binning on variable values of the sample characteristic variables aiming at any sample characteristic variable of a wind control sample user to obtain a plurality of bins of the sample characteristic variables; screening multiple sample characteristic variables according to each sub-box of each sample characteristic variable to obtain a target variable; aiming at any sub-box of any target variable, determining a coding value corresponding to the sub-box of the target variable according to a wind control evaluation label of a wind control sample user; and determining the coding value corresponding to the sub-box as the target characteristic data of the wind control sample user of which the variable value of the target variable falls into the sub-box.

408. Dividing the target characteristic data, the wind-controlled duration information and the wind-controlled evaluation labels according to a preset proportion to obtain training sample data and test sample data;

in this embodiment, training sample data and test sample data are generated according to target characteristic data of the wind control sample user and a wind control duration label and a wind control evaluation label obtained from the wind control duration information of the wind control sample user.

The training set is a learning sample data set, and a classifier is established by matching some parameters, that is, the target training data in the training sample data is used to train the machine learning model to determine the parameters of the machine learning model. The test set is used to test the resolving power, such as recognition rate, of the trained machine learning model. In this embodiment, the training face pictures may be divided according to a ratio of 9:1, that is, 90% of the training face pictures may be used as training sample data, and the remaining 10% of the training face pictures may be used as test sample data.

409. Inputting training sample data into a preset convolutional neural network model for training to obtain an original wind control model;

in this embodiment, the convolutional neural network model is a model obtained by combining a convolutional neural network model and a long and short term recurrent neural network model. It can be understood that the convolutional neural network-long and short recurrent neural network model is equivalent to a model formed by connecting the convolutional neural network with the long and short recurrent neural network model.

Convolutional Neural Networks (CNN)) are locally connected networks. Compared with a fully-connected network, the method has the greatest characteristics of local connectivity and weight sharing. For a certain pixel p in an image, the closer the pixel p is, the larger the influence (local connectivity) on it. In addition, according to the statistical characteristics of the natural image, the weight of a certain region can also be used for another region, i.e. the sharing of the weight. The weight sharing can be understood as convolution kernel sharing, in a Convolution Neural Network (CNN), one convolution kernel and a given image are subjected to convolution operation to extract one image feature, and different convolution kernels can extract different image features. Due to the local connectivity of the convolutional neural network, the complexity of the model is reduced, and the efficiency of model training is improved; and moreover, because of the weight sharing property of the convolutional neural network, the convolutional neural network can learn in parallel, and the model training efficiency is further improved.

410. Inputting test sample data into an original wind control model for testing to obtain a target wind control model;

in this embodiment, the target wind control model is a model in which the original risk model is tested by using a training face picture in test sample data, so that the accuracy of the original wind control model reaches a preset accuracy. Specifically, target training data in the test sample data, namely training face pictures of continuous N frames, are adopted to test the original wind control model so as to obtain corresponding accuracy; and if the accuracy reaches the preset accuracy, taking the original wind control model as a target wind control model.

411. And inputting the target service data and the risk assessment factor into a preset wind control model for analysis to obtain a risk assessment result of the service data.

The

steps

401, 404, 411 in this embodiment are similar to the

steps

101, 104, 105 in the first embodiment, and are not described herein again.

Referring to fig. 5, a fifth embodiment of the risk assessment method according to the embodiment of the present invention includes:

501. acquiring original service data, and cleaning the service data to obtain standard service data;

502. acquiring historical service data according to the data type of the standard service data, and configuring a wind control rule according to the historical service data;

503. performing dimension splitting on the service data according to a preset dimension to generate a plurality of target service data;

504. determining a wind control evaluation factor corresponding to the target service data according to a wind control rule;

505. acquiring sample characteristic data of a plurality of wind control sample users, wind control expression data of the wind control sample users in a scene to be wind controlled and wind controlled duration information of the wind control sample users in the scene to be wind controlled;

506. determining a wind control evaluation label of a wind control sample user according to the wind control performance data;

507. processing the sample characteristic data according to the wind control evaluation label to obtain target characteristic data of a wind control sample user;

508. dividing the target characteristic data, the wind-controlled duration information and the wind-controlled evaluation labels according to a preset proportion to obtain training sample data and test sample data;

509. initializing model parameters of a convolutional neural network model;

in this embodiment, initializing the convolutional neural network model refers to initializing the model parameters (i.e., the convolutional kernel and the offset) of the convolutional neural network model in advance. The convolution kernel is a weight of a convolution neural network, and when training data is input, the training data is multiplied by a weight, namely the convolution kernel, and then the output of a neuron is obtained, which reflects the importance degree of the training data. The bias is a linear component used to alter the range of weights multiplied by the input. And completing the process of model training based on the determined connection weight values between the layers in the convolution kernel and the bias.

510. Performing feature extraction on training sample data by adopting a convolutional neural network to obtain a feature vector of the training sample data;

in this embodiment, the feature vector is a facial feature obtained by performing feature extraction on target training data in a training set by using a convolutional neural network. Specifically, the method for extracting the features of the target training data in the training set by using the convolutional neural network comprises the following steps:

the feature vector is obtained by performing convolution operation on target training data in a training set by adopting a convolution neural network model. Specifically, the maximum pooling downsampling is adopted to perform downsampling operation on the convolved feature map so as to realize dimensionality reduction on the feature map, and the calculation formula is that yj represents the ith output spectrum (namely the downsampled feature map) in the downsampling process, and each neuron in the downsampling process is obtained by locally sampling the ith input spectrum (the convolved feature map) by adopting a downsampling frame of S & ltS & gt; m and n represent the step size of the down-sampling frame movement, respectively.

511. Inputting the feature vector of the training sample data into a convolutional neural network model for training to obtain an original wind control model;

in this embodiment, the LSTM model is one of neural network models with long-term memory capability, and has a three-layer network structure of an input layer, a hidden layer, and an output layer. The input layer is the first layer of the LSTM model and is used for receiving external signals, i.e., is responsible for receiving eigenvectors carrying time sequence states. In this embodiment, since the training sample data has a time sequence, the feature vector of the training sample data also has a time sequence, so that the training sample data can be applied to the LSTM model, and the LSTM can obtain the feature vector carrying the time sequence state. The output layer is the last layer of the LSTM model and is used for outputting signals to the outside, i.e. responsible for outputting the calculation results of the LSTM model. The hidden layer is a layer except the input layer and the output layer in the LSTM model and is used for processing the input feature vectors and obtaining the calculation result of the LSTM model.

The original wind control model is a model obtained by adopting an LSTM model to carry out multiple iterations on the characteristic vector carrying the time sequence state until convergence. Understandably, model training is performed on the extracted feature vectors by adopting an LSTM model, so that the time sequence of the obtained original wind control model is enhanced, and the accuracy of the original wind control model is improved.

In the embodiment, the convolutional neural network-long-and-short-term recurrent neural network model is initialized, so that the target training data in the training set is trained based on the convolutional neural network model to obtain the feature vectors, and then the obtained face features are input into the LSTM model for training.

512. Inputting test sample data into an original wind control model for testing to obtain a target wind control model;

513. and inputting the target service data and the risk assessment factor into a preset wind control model for analysis to obtain a risk assessment result of the service data.

The steps 501-504, 513 in this embodiment are similar to the steps 101-104, 105 in the first embodiment, and are not described herein again.

With reference to fig. 6, the risk assessment method in the embodiment of the present invention is described above, and a risk assessment apparatus in the embodiment of the present invention is described below, where a first embodiment of the risk assessment apparatus in the embodiment of the present invention includes:

a data cleaning module 601, configured to obtain original service data and clean the service data to obtain standard service data;

a configuration module 602, configured to obtain historical service data according to the data type of the standard service data, and configure a wind control rule according to the historical service data;

a generating module 603, configured to perform dimension splitting on the service data according to a preset dimension, and generate multiple target service data;

a first determining module 604, configured to determine, according to the wind control rule, a wind control evaluation factor corresponding to the target service data;

and the analysis module 605 is configured to input the target service data and the risk assessment factor into a preset wind control model for analysis, so as to obtain a risk assessment result of the service data.

Referring to fig. 7, a risk assessment apparatus according to a second embodiment of the present invention specifically includes:

In this embodiment, the data cleansing module 601 is specifically configured to:

carrying out duplication removal operation on the original service data, and detecting whether the original service data after duplication removal has a data missing value or not;

if the missing data value does not exist, the original service data after the duplication removal is used as standard service data;

and if the data missing value exists, performing data filling on the data missing value to obtain standard service data.

In this embodiment, the configuration module 602 includes:

an extracting unit 6021, configured to determine a type of the service data, obtain historical service data based on the type, extract a wind control analysis result in the historical service data, and determine a risk threshold;

an adjusting unit 6022, configured to adjust a preset wind control strategy according to the risk threshold to obtain an actual wind control strategy;

a determining unit 6023, configured to determine a wind control effect characterization value corresponding to the actual wind control strategy according to the actual wind control strategy and the wind control analysis result;

a configuration unit 6024, configured to configure a wind control rule according to the wind control effect representation value and the historical service data.

The configuration unit 6024 is specifically configured to:

determining a target wind control effect representation value corresponding to the wind control strategy according to the wind control strategy and the historical service data;

and taking the target wind control effect representation value as a test parameter, and configuring a wind control rule according to the target range and the historical service data.

In this embodiment, the generating module 603 is specifically configured to:

acquiring a data field of the standard service data, and identifying the data attribute of the standard service data according to the data field;

clustering the standard service data with the same data attribute to obtain a data clustering center point;

and generating corresponding target service data according to the data clustering center point.

In this embodiment, the risk assessment apparatus further includes:

an obtaining module 606, configured to obtain sample feature data of multiple wind control sample users, wind control performance data of the wind control sample users in a scene to be wind controlled, and wind controlled duration information of the wind control sample users in the scene to be wind controlled;

a second determining module 607, configured to determine, according to the wind control performance data, a wind control evaluation tag of the wind control sample user; processing the sample characteristic data according to the wind control evaluation label to obtain target characteristic data of the wind control sample user;

a dividing module 608, configured to divide the target feature data, the controlled duration information, and the wind control evaluation label according to a preset ratio to obtain training sample data and test sample data;

the training module 609 is configured to input the training sample data into a preset convolutional neural network model for training to obtain an original wind control model;

and the test module 610 is configured to input the test sample data into the original wind control model for testing, so as to obtain a target wind control model.

In this embodiment, the training module 609 is specifically configured to:

initializing model parameters of a convolutional neural network model;

performing feature extraction on the training sample data by adopting a convolutional neural network to obtain a feature vector of the training sample data;

and inputting the feature vector of the training sample data into the convolutional neural network model for training to obtain the original wind control model.

Fig. 6 and fig. 7 describe the risk assessment apparatus in the embodiment of the present invention in detail from the perspective of the modular functional entity, and the risk assessment apparatus in the embodiment of the present invention is described in detail from the perspective of hardware processing.

Fig. 8 is a schematic structural diagram of a risk assessment device according to an embodiment of the present invention, where the risk assessment device 800 may have a relatively large difference due to different configurations or performances, and may include one or more processors (CPUs) 810 (e.g., one or more processors) and a memory 820, and one or more storage media 830 (e.g., one or more mass storage devices) storing an application 833 or data 832. Memory 820 and storage medium 830 may be, among other things, transient or persistent storage. The program stored in the storage medium 830 may include one or more modules (not shown), each of which may include a series of instructions operating on the risk assessment device 800. Further, the processor 810 may be configured to communicate with the storage medium 830, and execute a series of instruction operations in the storage medium 830 on the risk assessment device 800 to implement the steps of the risk assessment method provided by the above-described method embodiments.

The risk assessment device 800 may also include one or more power supplies 840, one or more wired or wireless network interfaces 850, one or more input-output interfaces 860, and/or one or more operating systems 831, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, and the like. Those skilled in the art will appreciate that the configuration of the risk assessment device illustrated in FIG. 8 does not constitute a limitation of the risk assessment devices provided herein, and may include more or less components than those illustrated, or some components in combination, or a different arrangement of components.

The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium, and which may also be a volatile computer-readable storage medium, having stored therein instructions, which, when executed on a computer, cause the computer to perform the steps of the above-described risk assessment method.

It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.

The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims

1. A risk assessment method, characterized in that the risk assessment method comprises:

acquiring original service data, and cleaning the service data to obtain standard service data;

acquiring historical service data according to the data type of the standard service data, and configuring a wind control rule according to the historical service data;

performing dimension splitting on the service data according to a preset dimension to generate a plurality of target service data;

determining a wind control evaluation factor corresponding to the target service data according to the wind control rule;

and inputting the target service data and the risk assessment factor into a preset wind control model for analysis to obtain a risk assessment result of the service data.

2. The risk assessment method according to claim 1, wherein the obtaining of the original business data and the cleaning of the business data to obtain the standard business data comprises:

acquiring original service data, performing duplicate removal operation on the original service data, and detecting whether the original service data subjected to duplicate removal has a data missing value or not;

3. The risk assessment method according to claim 1, wherein the obtaining historical business data according to the data type of the standard business data and configuring the wind control rule according to the historical business data comprises:

determining the type of the service data, acquiring historical service data based on the type, extracting a wind control analysis result in the historical service data, and determining a risk threshold;

adjusting a preset wind control strategy according to the risk threshold value to obtain an actual wind control strategy;

determining a wind control effect representation value corresponding to the actual wind control strategy according to the actual wind control strategy and the wind control analysis result;

and configuring a wind control rule according to the wind control effect representation value and the historical service data.

4. The risk assessment method according to claim 3, wherein the configuring of the wind control rule according to the wind control effect representation value and the historical business data comprises:

determining a wind control effect representation value corresponding to the actual wind control strategy according to the actual wind control strategy and the historical service data;

and taking the wind control effect characteristic value as a test parameter, and configuring a wind control rule according to a target range corresponding to a preset wind control effect characteristic value and the historical service data.

5. The risk assessment method according to claim 1, wherein the performing dimension splitting on the business data according to a preset dimension to generate a plurality of target business data comprises:

6. The risk assessment method according to claim 1, wherein before the inputting the target business data and the risk assessment factor into a preset wind control model for analysis to obtain a risk assessment result of the business data, the method further comprises:

acquiring sample characteristic data of a plurality of wind control sample users, wind control performance data of the wind control sample users in a scene to be wind controlled and wind controlled duration information of the wind control sample users in the scene to be wind controlled;

determining a wind control evaluation label of the wind control sample user according to the wind control performance data;

processing the sample characteristic data according to the wind control evaluation label to obtain target characteristic data of the wind control sample user;

dividing the target characteristic data, the wind-controlled duration information and the wind-controlled evaluation label according to a preset proportion to obtain training sample data and test sample data;

inputting the training sample data into a preset convolutional neural network model for training to obtain an original wind control model;

and inputting the test sample data into the original wind control model for testing to obtain a target wind control model.

7. The risk assessment method according to claim 6, wherein the inputting the training sample data into a preset convolutional neural network model for training to obtain an original wind control model comprises:

initializing model parameters of a convolutional neural network model;

8. A risk assessment device, characterized in that it comprises:

the data cleaning module is used for acquiring original service data and cleaning the service data to obtain standard service data;

the configuration module is used for acquiring historical service data according to the data type of the standard service data and configuring a wind control rule according to the historical service data;

the generating module is used for carrying out dimension splitting on the service data according to a preset dimension to generate a plurality of target service data;

the first determining module is used for determining a wind control evaluation factor corresponding to the target service data according to the wind control rule;

and the analysis module is used for inputting the target service data and the risk assessment factor into a preset wind control model for analysis to obtain a risk assessment result of the service data.

9. A risk assessment device, characterized in that it comprises: a memory having instructions stored therein and at least one processor, the memory and the at least one processor interconnected by a line;

the at least one processor invoking the instructions in the memory to cause the risk assessment device to perform the steps of the risk assessment method of any one of claims 1-7.

10. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the risk assessment method according to any one of claims 1-7.