CN116151965B - Risk feature extraction method and device, electronic equipment and storage medium - Google Patents
Risk feature extraction method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN116151965B CN116151965B CN202310348690.9A CN202310348690A CN116151965B CN 116151965 B CN116151965 B CN 116151965B CN 202310348690 A CN202310348690 A CN 202310348690A CN 116151965 B CN116151965 B CN 116151965B
- Authority
- CN
- China
- Prior art keywords
- user
- risk
- behavior
- credit
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/95—Pattern authentication; Markers therefor; Forgery detection
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The application provides a risk feature extraction method, a risk feature extraction device, electronic equipment and a storage medium, wherein the risk feature extraction method comprises the following steps: acquiring user probe data and a user identification image; extracting user behavior characteristics according to the user probe data; extracting user identity characteristics according to the user identity identification image; the risk features are obtained by aggregating user behavioral features, user identity features and credit behavioral features. The method comprises the steps of aggregating user probe data and user identification data for fraud identification with credit behavior characteristics for credit admission credit, obtaining aggregated risk characteristics for fraud identification and/or credit admission credit, effectively aggregating fraud identification and credit admission credit together, realizing integrated processing of fraud identification and credit admission credit, and improving credit approval business efficiency.
Description
Technical Field
The present disclosure relates to the field of computer technologies, and in particular, to a risk feature extraction method, a risk feature extraction device, an electronic device, and a storage medium.
Background
With the rapid development of online retail credit services, risk prevention and control face more complex challenges, and the identification of traditional credit risks is changed from original to identification of emerging varied fraud risks and multi-customer group credit risks, and model strategies of fraud risks and credit risks are two relatively independent islands.
The fraud identification method and the credit admission credit authorization method in the related technology are two independent methods, and in the business which needs to carry out fraud identification and credit admission credit authorization on the user, the efficiency of carrying out business approval by using the two independent methods is lower.
Disclosure of Invention
An objective of the embodiments of the present application is to provide a risk feature extraction method, a risk feature extraction device, an electronic device, and a storage medium, which are used for approval efficiency of credit business.
In a first aspect, an embodiment of the present application provides a risk feature extraction method, including: acquiring user probe data and a user identification image; extracting user behavior characteristics according to the user probe data; extracting user identity characteristics according to the user identity identification image; and aggregating the user behavior characteristics, the user identity characteristics and the credit behavior characteristics to obtain risk characteristics.
In an implementation manner of the first aspect, the extracting a user behavior feature according to the user probe data includes: extracting user behavior characteristics by adopting a behavior characteristic extraction model according to the user probe data;
the training method of the behavior feature extraction model comprises the following steps: training the behavior feature extraction model by adopting first training data with a fraud tag until the training stopping condition of the behavior feature extraction model is met; acquiring the user behavior characteristics extracted by the behavior characteristic extraction model; calculating correlation coefficients among the user behavior characteristics; and adjusting model parameters of the behavior feature extraction model according to the correlation coefficient and a preset aggregation degree adjustment strategy, and retraining the behavior feature extraction model after the model parameters are adjusted.
In the implementation process of the scheme, the trained behavior feature extraction model is adopted to automatically extract the behavior features of the user, so that the problems of sparsity and low effectiveness caused by the existing structural derivatization are effectively solved, and the effectiveness of the aggregated risk features is improved; meanwhile, the feature extraction model can adjust the aggregation degree of the features according to the correlation coefficient between the extracted user behavior features during training, so that the risk feature extraction method can adapt to more scenes, and the adaptability of the risk feature extraction method is improved.
In an implementation manner of the first aspect, the aggregating the user behavior feature, the user identity feature and the credit behavior feature to obtain a risk feature includes: splicing the user behavior characteristic and the user identity characteristic to obtain a first splicing characteristic; calculating an IV value of the first splicing characteristic under a preset fraud label; screening the first splicing characteristics meeting a preset IV threshold value to obtain second splicing characteristics; and splicing the second splicing characteristic and the credit behavior characteristic to obtain a risk characteristic.
In the implementation process of the scheme, the user behavior characteristics and the user identity characteristics which can be effectively identified in a fraud manner are screened out through the IV value of the first splicing characteristics under the fraud label, and then the effective characteristics and the credit behavior characteristics are spliced, so that the effectiveness of the risk characteristics is realized.
In an implementation manner of the first aspect, after the aggregating the user behavior feature, the user identity feature and the credit behavior feature to obtain a risk feature, the method further includes: an interpretable description of the risk feature is generated.
In the implementation process of the scheme, if the risk features are adopted to perform fraud recognition or credit admission authorization, the obtained recognition result or admission authorization result is correspondingly provided with the explanatory description, so that the credibility of the risk features extracted by the risk feature extraction method is improved.
In an implementation manner of the first aspect, the extracting a user behavior feature according to the user probe data includes: and extracting at least one of event sequence characteristics, touch behavior characteristics and physical signal characteristics according to the user probe data.
In the implementation process of the scheme, the user behavior features which can be extracted according to the user probe data comprise at least one of event sequence features, touch behavior features and physical signal features, and specific feature combinations can be selected according to actual requirements, so that the risk feature extraction method can be suitable for more scenes, and the adaptability of the risk feature extraction method is improved.
The risk feature extraction method has the beneficial effects that: the method comprises the steps of aggregating user probe data and user identification data for fraud identification with credit behavior characteristics for credit admission credit, obtaining aggregated risk characteristics for fraud identification and/or credit admission credit, effectively aggregating fraud identification and credit admission credit together, realizing integrated processing of fraud identification and credit admission credit, and improving credit approval business efficiency; meanwhile, the fraud identification precision and credit scoring accuracy are improved, and the false detection rate of credit business approval is reduced.
In a second aspect, an embodiment of the present application provides a fraud risk identification method, where the fraud risk identification method uses risk features extracted by any one of the risk feature extraction methods to perform fraud risk identification, including: calculating a risk characteristic value according to user probe data of a user to be identified and a user identification image; and determining the fraud risk of the user to be identified according to the risk characteristic value and the preset risk characteristic threshold value of each fraud label.
In a third aspect, an embodiment of the present application provides a credit admission credit granting method, where the method performs credit admission credit granting by using risk features extracted by any one of the risk feature extraction methods, including: calculating a risk characteristic value according to user probe data and a user identification image of a user to be trusted; inputting the risk characteristics and the risk characteristic values into a credit score model to obtain credit scores of the users to be trusted; and determining the admission result and the credit limit of the user to be trusted according to the credit score.
In a fourth aspect, an embodiment of the present application provides a risk feature extraction apparatus, including:
the data acquisition module is used for acquiring user probe data and user identification images;
the behavior feature extraction module is used for extracting user behavior features according to the user probe data;
the identity feature extraction module is used for extracting the identity features of the user according to the user identity identification image;
and the feature aggregation module is used for aggregating the user behavior features, the user identity features and the credit behavior features to obtain risk features.
In a fifth aspect, embodiments of the present application provide a computer readable storage medium having stored thereon computer program instructions which, when read and executed by a processor, perform the method provided by the first aspect or any one of the possible implementations of the second aspect or any one of the possible implementations of the third aspect.
In a sixth aspect, an embodiment of the present application provides an electronic device, including: a memory and a processor, the memory having stored therein computer program instructions which, when read and executed by the processor, perform the method of the first aspect or any one of the possible implementations of the second aspect or any one of the possible implementations of the third aspect.
Additional features and advantages of the application will be set forth in the description which follows, and in part will be apparent from the description, or may be learned by practice of the embodiments of the application. The objectives and other advantages of the application will be realized and attained by the structure particularly pointed out in the written description and claims thereof as well as the appended drawings.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments of the present application will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic flow chart of a risk feature extraction method provided in an embodiment of the present application;
fig. 2 is a schematic structural diagram of a risk feature extraction device provided in an embodiment of the present application;
fig. 3 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application. The following examples are only for more clearly illustrating the technical solutions of the present application, and thus are only examples, and are not intended to limit the scope of protection of the present application.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having" and any variations thereof in the description and claims of the present application and in the description of the figures above are intended to cover non-exclusive inclusions.
In the description of the embodiments of the present application, the technical terms "first," "second," etc. are used merely to distinguish between different objects and are not to be construed as indicating or implying a relative importance or implicitly indicating the number of technical features indicated, a particular order or a primary or secondary relationship. In the description of the embodiments of the present application, the meaning of "plurality" is two or more unless explicitly defined otherwise.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the present application. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments.
In the description of the embodiments of the present application, the term "and/or" is merely an association relationship describing an association object, which means that three relationships may exist, for example, a and/or B may mean: a exists alone, A and B exist together, and B exists alone. In addition, the character "/" herein generally indicates that the front and rear associated objects are an "or" relationship.
Referring to fig. 1, an embodiment of the present application provides a risk feature extraction method, including:
step S110: acquiring user probe data and a user identification image;
step S120: extracting user behavior characteristics according to the user probe data;
step S130: extracting user identity characteristics according to the user identity identification image;
step S140: and aggregating the user behavior characteristics, the user identity characteristics and the credit behavior characteristics to obtain risk characteristics.
In the implementation process of the scheme, the user probe data and the user identification data for fraud identification and the credit behavior characteristics for credit admission credit are aggregated, the aggregated risk characteristics are obtained to carry out fraud identification and/or credit admission credit, the fraud identification and the credit admission credit are effectively aggregated together, the integrated processing of fraud identification and credit admission credit is realized, and the efficiency of credit approval business is improved.
Step S110 is described in detail below:
the user probe data acquired in step S110 refers to user data acquired using the bio-probe technology. Among them, the biological probe technique refers to: the intelligent terminal equipment is used for collecting biological characteristics of a user such as fingerprints, sound waves, faces, irises and the like, biological behavior characteristics of the user such as operation habits of the user and equipment information of the intelligent terminal equipment.
The user probe data acquired in the embodiment includes page event behavior, acceleration sensing, gyroscope sensing, sliding and rolling behavior, page touch behavior, equipment information and the like.
The method for acquiring the user probe data in step S110 may be: embedding points in an applet or APP terminal in advance, and then collecting embedded point data when a user uses the applet or APP terminal.
The user identification image acquired in step S110 includes: a live image and/or an identification card front image. The method for acquiring the living body image can be as follows: the living body image of the user is acquired at the time of the face-examination of the user. The method for acquiring the front image of the identity card can be as follows: the user automatically uploads the front image of the identity card.
It is understood that after step S110, it may further include: preprocessing the user probe data and the user identification image acquired in step S110, and the following processing methods for the user probe data and the user identification image are respectively described:
the processing method for the user probe data comprises the following steps: the user probe data acquired in step S110 is divided into:
event sequence data including loading times, equipment brands, GPS longitude and latitude, equipment ID, equipment OS, equipment network status, etc.;
and/or touch behavior data including sliding speed, sliding distance, sliding duration, touch times, touch positions, touch duration, etc.;
and/or physical signal data, acceleration sensing data, gyroscope sensing data, and the like.
It should be noted that, when processing the user probe data, the user probe data needs to be further subjected to blocking processing, and the blocking rule may be a timing blocking or an event triggering blocking, where the event triggering blocking includes: when a user is in a certain page, the collected event sequence data, touch behavior data and physical signal data are packaged into one data block, and when the user jumps to another page, the collected event sequence data, touch behavior data and physical signal data are packaged into another data block, and specific partitioning rules can be selected according to specific application requirements.
After classifying and blocking the user probe data, normalizing the user probe data.
The processing method for the user identification image comprises the following steps: and carrying out standardization processing on the user identification image.
Step S120 is described in detail below:
optionally, step S120 extracts user behavior features according to the user probe data, including: and extracting the behavior characteristics of the user by adopting a behavior characteristic extraction model according to the user probe data.
The training method of the behavior feature extraction model comprises the following steps:
training the behavior feature extraction model by adopting first training data with fraud labels such as information counterfeiting, decoupling and the like until training stop conditions of the behavior feature extraction model are met, wherein the training stop conditions can be that a loss function of the behavior feature extraction model reaches a preset loss threshold requirement and the iteration times reach a preset time requirement;
acquiring the user behavior characteristics extracted by the behavior characteristic extraction model;
calculating a correlation coefficient between user behavior features;
and adjusting model parameters of the behavior feature extraction model according to the correlation coefficient and a preset aggregation degree adjustment strategy to adjust the aggregation degree of the features, and retraining the behavior feature extraction model after the adjustment of the model parameters until the aggregation degree requirement is met.
Optionally, the user behavior feature comprises at least one of an event sequence feature, a touch behavior feature, and a physical signal feature.
It can be understood that the event sequence feature, the touch behavior feature and the physical signal feature can be extracted by respectively adopting the event sequence data, the touch behavior data and the physical signal data, and the extraction method is as follows:
the behavioral characteristic extraction model may include: a BiLSTM submodel and a transducer submodel;
extracting event sequence features by adopting a BiLSTM sub-model and event sequence data;
extracting touch behavior characteristics by adopting a BiLSTM sub-model and touch behavior data;
and extracting physical signal characteristics by using a transducer submodel and physical signal data.
It should be noted that when the BiLSTM submodel and the transducer submodel are used to extract the event sequence feature, the touch behavior feature and the physical signal feature, the last layer of the original model structure needs to be modified into the embedding layer, so that the event sequence feature, the touch behavior feature and the physical signal feature can be obtained through the embedding layer.
It should be noted that, when the above-mentioned BiLSTM sub-model and the above-mentioned transducer sub-model extract the event sequence feature, the touch behavior feature and the physical signal feature, training is required respectively, and the feature aggregation degree is adjusted according to the correlation coefficient between the features extracted by the model after training is completed, so that the above-mentioned event sequence feature, touch behavior feature and physical signal feature all meet the feature aggregation degree requirement.
It should be noted that: the method for generating the interpretable description of the user behavior features in the model training stage can be as follows:
after the user behavior features extracted by the behavior feature extraction model meet the feature aggregation degree requirements, calculating IV values of the user behavior features under the fraud labels, and screening effective user behavior features according to a preset IV value threshold;
visual analysis is carried out on the user behaviors of the polar cases according to the quantiles of the values of the user behavior characteristics, such as 5% quantiles and 95% quantiles;
and (3) acquiring the interpretable description of the behavior characteristics of the user according to expert experience, such as frequent heavy test, page jerky click sliding, abnormal mobile phone holding gesture and the like.
Step S130 is described in detail below:
step S130 may employ an identity feature extraction model, such as the res net16 model, to extract user identity features based on the fraud tag.
It should be noted that, in the model training stage, the interpretable description of the user identity feature may be generated, so that the user identity feature extracted by using the identity feature extraction model also has the interpretable description, and in the model training stage, the method for generating the interpretable description of the user identity feature may be as follows:
acquiring a focus point of the identity feature extraction model on an image based on a shape algorithm and a Grad-CAM algorithm, and performing visual processing on the focus point; the shape algorithm adopts an optimal information distribution and local interpretation algorithm, obtains a shape value by calculating a joint arrangement average value, and visualizes the shape value by thermodynamic diagram; the Grad-CAM algorithm calculates weight values by utilizing gradients, adopts weight scores generated by a global average pooling method, applies weight values of a full-connection layer, and visualizes the weight values through thermodynamic diagrams;
and acquiring the interpretable description of the user identity characteristics such as the image key points, the living body contours, the background information and the like according to the visualized image attention points.
Step S140 is described in detail below:
optionally, step S140 aggregates the user behavior feature, the user identity feature and the credit behavior feature to obtain the risk feature, including:
splicing the user behavior characteristic and the user identity characteristic to obtain a first spliced characteristic;
calculating an IV value of the first splicing characteristic under a preset fraud label;
screening the first splicing characteristic meeting the preset IV threshold value to obtain a second splicing characteristic;
and splicing the second splicing characteristic with the credit behavior characteristic to obtain the risk characteristic.
The credit behavior feature refers to: credit card liability features, loan summary features, personal information summary features, multi-headed co-liability features, etc. features for credit approval, which also have an interpretable description.
The method for splicing the user behavior feature and the user identity feature and the first splicing feature and the credit behavior feature may be: and directly splicing and combining the features.
Optionally, step S140 further includes, after aggregating the user behavior feature, the user identity feature and the credit behavior feature to obtain the risk feature: an interpretable description of the risk feature is generated.
Since step S120 and step S130 generate an interpretable description of the feature when extracting the user behavior feature and the user identity feature, respectively, and the credit behavior feature also has an interpretable description, the risk feature generated by aggregating the user behavior feature, the user identity feature, and the credit behavior feature also has an interpretable description.
It should be noted that, the embodiment of the application also judges the credibility of the risk features to fraud identification and credit admission credit authorization services by the following method:
performing univariate analysis on the risk characteristics, and screening the characteristics with strong stability and higher IV value;
putting the screened features into a decision tree model, training the model based on credit labels, and outputting feature importance;
drawing a tree structure through Graphviz;
based on the drawn decision tree, feature ordering between the root node and the partition node of the tree is mined, and if the feature is positioned at the root node of the decision tree, the importance of the feature is higher and becomes an important condition for client grouping;
through analysis of risk characteristics, the user behavior characteristics and the user identity characteristics constructed based on the anti-fraud purpose are found to be root nodes on credit labels, namely credit scoring can be realized for guest groups which do not pass through fraud labels, and each branch of the constructed decision tree and the GINI coefficient can be optimized. From this it follows that: the user behavior features and user identity features for fraud identification have an association with credit behavior features for credit admission credit.
In addition, a control experiment is further provided in the embodiment of the application to verify the effectiveness of the risk features extracted by the risk feature extraction method, and the control experiment is as follows:
setting an experimental group, a control group 1 and a control group 2 which adopt the same training set and verification set, wherein:
the setting method of the experimental group comprises the following steps: training the XGBoost model based on a single overdue task by adopting risk characteristics, searching through a network to reach optimal parameters, and completing multiple iterations of the model until the performance is maximized to realize stability;
the method for setting the control group 1 is as follows: adopting a traditional machine learning algorithm, applying structural features such as equipment features, basic features, credit behavior features and the like, training and optimizing an XGBoost model based on a single task and a task thereof, and obtaining a model result;
the method for setting the control group 2 is as follows: and extracting images, event sequence features, touch behavior features and physical signal features based on the single overdue task, and completing training of the XGBoost model to obtain a model result.
The experimental group, the control group 1 and the control group 2 were tested on the training set and the test set, and a model KS (Kolmogorov-Smirnov) index, a model AUC (Area Under Curve) index, a model Accuracy index, a model Precision index and a model Recall index were obtained, and the results are shown in table 1.
Table 1 performance comparison table
As can be seen from table 1, the test results of the experimental group are better than those of the control group 2, and the test results of the control group 2 are better than those of the control group 1. From the test set performance, control group 1 had a test set ks of only 0.33, well below 0.54 of the experimental group, and control group 2, while incorporating high dimensional data, was also lower in performance than the experimental group based on a single task implementation.
Based on the same inventive concept, the application implementation also provides a fraud risk identification method, which adopts the risk features extracted by the risk feature extraction method to perform fraud risk identification, and comprises the following steps:
calculating a risk characteristic value according to user probe data of a user to be identified and a user identification image;
and determining the fraud risk of the user to be identified according to the risk characteristic value and the preset risk characteristic threshold value of each fraud label.
It will be appreciated that the risk feature thresholds for the various fraud tags may be preset and then the fraud tag of the user to be identified is determined from the calculated risk feature values, thereby determining the risk of fraud for the user to be identified.
Based on the same inventive concept, the application implementation also provides a credit admission credit granting method, which adopts the risk features extracted by the risk feature extraction method to perform credit admission credit granting, and comprises the following steps:
calculating a risk characteristic value according to user probe data and a user identification image of a user to be trusted;
inputting the risk characteristics and the risk characteristic values into a credit scoring model to obtain credit scores of users to be trusted;
and determining the admission result and the credit limit of the user to be trusted according to the credit score.
It should be noted that the credit scoring model may be a trained XGBoost model.
Compared with a linear model established based on structural features derived from expert experience and adopted by credit risk prevention and control in the related technology, the risk feature extraction method completes extraction of key information in a feature aggregation mode, and solves the problems of sparse distribution, poor stability and low effectiveness of the existing structural features. Meanwhile, the risk characteristics can be put into the tree model to serve as a root node, so that the condition of grouping guest groups is realized, then the credit qualification ordering of different fraud groups is realized according to the splitting of leaf child nodes, and the anti-fraud and credit admission credit granting are effectively combined together.
Based on the same inventive concept, the embodiment of the application also provides an integrated identification method for fraud risk and credit admission trust, comprising the following steps:
calculating a risk characteristic value according to user probe data of a user to be identified and a user identification image;
determining the fraud risk of the user to be identified according to the risk characteristic value and the preset risk characteristic threshold value of each fraud label;
inputting the risk characteristics and the risk characteristic values into a credit scoring model to obtain credit scores of users to be trusted;
and determining the admission result and the credit limit of the user to be trusted according to the credit score.
Referring to fig. 2, based on the same inventive concept, the present application further provides a risk feature extraction apparatus 200, including:
a data acquisition module 210 for acquiring user probe data and user identification images;
a behavior feature extraction module 220, configured to extract user behavior features according to the user probe data;
an identity feature extraction module 230, configured to extract user identity features according to the user identity document image;
and the feature aggregation module 240 is configured to aggregate the user behavior feature, the user identity feature and the credit behavior feature to obtain a risk feature.
Optionally, the behavior feature extraction module 220 specifically is: and extracting the behavior characteristics of the user by adopting a behavior characteristic extraction model according to the user probe data.
Optionally, the risk feature extraction device 200 further includes:
the behavior feature extraction model training module is used for training the behavior feature extraction model by adopting first training data with fraud labels until the training stopping condition of the behavior feature extraction model is met;
the aggregation degree adjusting module is used for acquiring the user behavior characteristics extracted by the behavior characteristic extraction model; calculating correlation coefficients among the user behavior characteristics; and adjusting model parameters of the behavior feature extraction model according to the correlation coefficient and a preset aggregation degree adjustment strategy, and retraining the behavior feature extraction model after the model parameters are adjusted.
Optionally, the feature aggregation module 240 includes:
the first splicing unit is used for splicing the user behavior characteristics and the user identity characteristics to obtain first splicing characteristics;
the feature screening unit is used for calculating the IV value of the first splicing feature under a preset fraud label, screening the first splicing feature meeting a preset IV threshold value and obtaining a second splicing feature;
and the second splicing unit is used for splicing the second splicing characteristic and the credit behavior characteristic and acquiring the risk characteristic.
Optionally, the risk feature extraction device 200 further includes:
and the interpretability description generation unit is used for generating an interpretability description of the risk characteristic.
Optionally, the behavior feature extraction module 220 includes:
at least one of an event sequence feature extraction unit, a touch behavior feature extraction unit and a physical signal feature extraction unit.
Based on the same inventive concept, the application implementation further provides a fraud risk identification apparatus, including:
the first risk characteristic value calculation module is used for calculating a risk characteristic value according to user probe data of a user to be identified and a user identification image;
and the fraud risk judging module is used for determining the fraud risk of the user to be identified according to the risk characteristic value and the preset risk characteristic threshold value of each fraud label.
Based on the same inventive concept, the application implementation also provides a credit admission credit device, which comprises:
the second risk characteristic value calculation module is used for calculating a risk characteristic value according to the user probe data and the user identification image of the user to be trusted;
the credit score obtaining unit is used for inputting the risk characteristics and the risk characteristic values into a credit score model to obtain the credit score of the user to be trusted;
and the admission credit determining unit is used for determining the admission result and credit limit of the user to be trusted according to the credit score.
Fig. 3 is a schematic diagram of an electronic device according to an embodiment of the present application. Referring to fig. 3, the electronic device 300 includes: processor 310, memory 320, and communication interface 330, which are interconnected and communicate with each other by a communication bus 340 and/or other forms of connection mechanisms (not shown).
The Memory 320 includes one or more (Only one is shown in the figure), which may be, but is not limited to, random Access Memory (RAM), read Only Memory (ROM), programmable Read Only Memory (Programmable Read-Only Memory, PROM), erasable Read Only Memory (Erasable Programmable Read-Only Memory, EPROM), electrically erasable Read Only Memory (Electric Erasable Programmable Read-Only Memory, EEPROM), and the like. The processor 310, as well as other possible components, may access, read, and/or write data from, the memory 320.
The processor 310 includes one or more (only one shown) which may be an integrated circuit chip having signal processing capabilities. The processor 310 may be a general-purpose processor, including a Central Processing Unit (CPU), a micro control unit (Micro Controller Unit MCU), a Network Processor (NP), or other conventional processors; but may also be a special purpose processor including a Digital Signal Processor (DSP), an application specific integrated circuit (Application Specific Integrated Circuits ASIC), a field programmable gate array (Field Programmable Gate Array FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components.
The communication interface 330 includes one or more (only one shown) that may be used to communicate directly or indirectly with other devices for data interaction. For example, the communication interface 330 may be an ethernet interface; may be a mobile communications network interface, such as an interface of a 3G, 4G, 5G network; or may be other types of interfaces with data transceiving functionality.
One or more computer program instructions may be stored in memory 320 that may be read and executed by processor 310 to implement the risk feature extraction method or fraud risk identification method or credit admission authorization method provided by embodiments of the present application, as well as other desired functions.
It is to be understood that the configuration shown in fig. 3 is illustrative only, and that electronic device 300 may also include more or fewer components than shown in fig. 3, or have a different configuration than shown in fig. 3. The components shown in fig. 3 may be implemented in hardware, software, or a combination thereof. For example, the electronic device 300 may be a single server (or other device having computing processing capabilities), a combination of multiple servers, a cluster of a large number of servers, or the like, and may be either a physical device or a virtual device.
The embodiment of the application also provides a computer readable storage medium, and the computer readable storage medium stores computer program instructions, and when the computer program instructions are read and run by a processor of a computer, the risk feature extraction method or the fraud risk identification method or the credit admission credit granting method provided by the embodiment of the application is executed. For example, the computer-readable storage medium may be implemented as memory 320 in electronic device 300 in FIG. 3.
In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. The above-described apparatus embodiments are merely illustrative, for example, the division of the units is merely a logical function division, and there may be other manners of division in actual implementation, and for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some communication interface, device or unit indirect coupling or communication connection, which may be in electrical, mechanical or other form.
Further, the units described as separate units may or may not be physically separate, and units displayed as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
Furthermore, functional modules in various embodiments of the present application may be integrated together to form a single portion, or each module may exist alone, or two or more modules may be integrated to form a single portion.
The foregoing is merely exemplary embodiments of the present application and is not intended to limit the scope of the present application, and various modifications and variations may be suggested to one skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.
Claims (9)
1. A risk feature extraction method, comprising:
acquiring user probe data and a user identification image;
extracting user behavior characteristics according to the user probe data; the user probe data includes: at least one of event sequence data, touch behavior data, and physical signal data;
extracting user identity characteristics according to the user identity identification image;
aggregating the user behavior characteristics, the user identity characteristics and credit behavior characteristics to obtain risk characteristics;
the extracting the user behavior characteristics according to the user probe data comprises the following steps: extracting user behavior characteristics by adopting a behavior characteristic extraction model according to the user probe data;
the step of extracting the user identity features according to the user identity image comprises the following steps: extracting user identity characteristics by adopting an identity characteristic extraction model according to the user identity identification image;
the step of aggregating the user behavior feature, the user identity feature and the credit behavior feature to obtain a risk feature comprises the following steps:
splicing the user behavior characteristic and the user identity characteristic to obtain a first splicing characteristic;
calculating an IV value of the first splicing characteristic under a preset fraud label;
screening the first splicing characteristics meeting a preset IV threshold value to obtain second splicing characteristics;
and splicing the second splicing characteristic and the credit behavior characteristic to obtain a risk characteristic.
2. The risk feature extraction method according to claim 1, wherein the training method of the behavioral feature extraction model includes:
training the behavior feature extraction model by adopting first training data with a fraud tag until the training stopping condition of the behavior feature extraction model is met;
acquiring the user behavior characteristics extracted by the behavior characteristic extraction model;
calculating correlation coefficients among the user behavior characteristics;
and adjusting model parameters of the behavior feature extraction model according to the correlation coefficient and a preset aggregation degree adjustment strategy, and retraining the behavior feature extraction model after the model parameters are adjusted.
3. The risk feature extraction method according to any one of claims 1-2, characterized in that after the aggregation of the user behavior feature, the user identity feature and the credit behavior feature, the risk feature is obtained, further comprising:
an interpretable description of the risk feature is generated.
4. The risk feature extraction method according to any one of claims 1-2, wherein the extracting the user behavior feature according to the user probe data includes:
and extracting at least one of event sequence characteristics, touch behavior characteristics and physical signal characteristics according to the user probe data.
5. A fraud risk identification method, characterized in that the fraud risk identification is performed by using the risk features extracted by the risk feature extraction method according to any one of claims 1 to 4, including:
calculating a risk characteristic value according to user probe data of a user to be identified and a user identification image;
determining the fraud risk of the user to be identified according to the risk characteristic value and a preset risk characteristic threshold value of each fraud label;
the determining the fraud risk of the user to be identified according to the risk characteristic value and the preset risk characteristic threshold value of each fraud tag comprises the following steps:
determining the fraud tag of the user to be identified according to the risk characteristic value and the preset risk characteristic threshold value of each fraud tag;
and determining the fraud risk of the user to be identified according to the fraud label of the user to be identified.
6. A credit admission credit granting method, characterized in that the credit admission credit granting is performed by adopting the risk features extracted by the risk feature extraction method according to any one of claims 1 to 4, comprising:
calculating a risk characteristic value according to user probe data and a user identification image of a user to be trusted;
inputting the risk characteristics and the risk characteristic values into a credit score model to obtain credit scores of the users to be trusted; the credit scoring model includes: an XGBoost model;
and determining the admission result and the credit limit of the user to be trusted according to the credit score.
7. A risk feature extraction device, comprising:
the data acquisition module is used for acquiring user probe data and user identification images;
the behavior feature extraction module is used for extracting user behavior features according to the user probe data; the user probe data includes: at least one of event sequence data, touch behavior data, and physical signal data;
the identity feature extraction module is used for extracting the identity features of the user according to the user identity identification image;
the feature aggregation module is used for aggregating the user behavior features, the user identity features and the credit behavior features to obtain risk features;
the behavior feature extraction module is specifically configured to: extracting user behavior characteristics by adopting a behavior characteristic extraction model according to the user probe data;
the identity characteristic extraction module is specifically configured to: extracting user identity characteristics by adopting an identity characteristic extraction model according to the user identity identification image;
the feature aggregation module comprises:
the first splicing unit is used for splicing the user behavior characteristics and the user identity characteristics to obtain first splicing characteristics;
the feature screening unit is used for calculating the IV value of the first splicing feature under a preset fraud label, screening the first splicing feature meeting a preset IV threshold value and obtaining a second splicing feature;
and the second splicing unit is used for splicing the second splicing characteristic and the credit behavior characteristic and acquiring the risk characteristic.
8. An electronic device, comprising: the device comprises a processor, a memory and a communication bus, wherein the processor and the memory complete communication with each other through the communication bus; the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the method of any of claims 1-4 or claim 5 or claim 6.
9. A non-transitory computer readable storage medium storing computer instructions which, when executed by a computer, cause the computer to perform the method of any one of claims 1-4 or claim 5 or claim 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310348690.9A CN116151965B (en) | 2023-04-04 | 2023-04-04 | Risk feature extraction method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310348690.9A CN116151965B (en) | 2023-04-04 | 2023-04-04 | Risk feature extraction method and device, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116151965A CN116151965A (en) | 2023-05-23 |
| CN116151965B true CN116151965B (en) | 2023-07-18 |
Family
ID=86339137
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310348690.9A Active CN116151965B (en) | 2023-04-04 | 2023-04-04 | Risk feature extraction method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116151965B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117078403B (en) * | 2023-10-18 | 2024-01-12 | 杭银消费金融股份有限公司 | Wind control decision method and system based on rule combination optimization |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110119980A (en) * | 2019-04-23 | 2019-08-13 | 北京淇瑀信息科技有限公司 | A kind of anti-fraud method, apparatus, system and recording medium for credit |
| CN115186973A (en) * | 2022-05-27 | 2022-10-14 | 民生科技有限责任公司 | Credit line assessment method and cascade system based on multi-platform combined modeling |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2014233006B2 (en) * | 2013-03-15 | 2017-06-15 | Socure Inc. | Risk assessment using social networking data |
| CN108985917A (en) * | 2018-06-05 | 2018-12-11 | 银谷普惠信息咨询（北京）有限公司 | For the operation processing method and device of loan application, mobile terminal |
| CN109118247A (en) * | 2018-08-01 | 2019-01-01 | 安徽好车到家信息科技有限公司 | The anti-fake system of automobile consumption credit and analysis method based on LBS database and customer portrait |
| CN109389494B (en) * | 2018-10-25 | 2021-11-05 | 北京芯盾时代科技有限公司 | Loan fraud detection model training method, loan fraud detection method and device |
| CA3132055A1 (en) * | 2019-03-01 | 2020-09-10 | Mastercard Technologies Canada ULC | Online application origination (oao) service for fraud prevention systems |
| CN111369342B (en) * | 2020-03-05 | 2023-10-17 | 中国建设银行股份有限公司 | Loan approval method, device, equipment and storage medium based on machine learning |
| CN112381644A (en) * | 2020-12-01 | 2021-02-19 | 天翼电子商务有限公司 | Credit scene risk user assessment method based on space variable reasoning |
| CN115392937A (en) * | 2022-10-25 | 2022-11-25 | 成都新希望金融信息有限公司 | User fraud risk identification method and device, electronic equipment and storage medium |
-
2023
- 2023-04-04 CN CN202310348690.9A patent/CN116151965B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110119980A (en) * | 2019-04-23 | 2019-08-13 | 北京淇瑀信息科技有限公司 | A kind of anti-fraud method, apparatus, system and recording medium for credit |
| CN115186973A (en) * | 2022-05-27 | 2022-10-14 | 民生科技有限责任公司 | Credit line assessment method and cascade system based on multi-platform combined modeling |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116151965A (en) | 2023-05-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111241989B (en) | Image recognition method and device and electronic equipment | |
| JP7414901B2 (en) | Living body detection model training method and device, living body detection method and device, electronic equipment, storage medium, and computer program | |
| CN110705428B (en) | Facial age recognition system and method based on impulse neural network | |
| CN110287813A (en) | Personal identification method and system | |
| CN116151965B (en) | Risk feature extraction method and device, electronic equipment and storage medium | |
| CN111291817A (en) | Image recognition method and device, electronic equipment and computer readable medium | |
| CN112668482B (en) | Face recognition training method, device, computer equipment and storage medium | |
| CN105740808A (en) | Human face identification method and device | |
| CN108875509A (en) | Biopsy method, device and system and storage medium | |
| CN112364803A (en) | Living body recognition auxiliary network and training method, terminal, equipment and storage medium | |
| CN111898561A (en) | Face authentication method, device, equipment and medium | |
| CN116311370A (en) | Multi-angle feature-based cow face recognition method and related equipment thereof | |
| CN111476668A (en) | Identification method and device of credible relationship, storage medium and computer equipment | |
| CN112819486B (en) | Method and system for identity certification | |
| CN116563898A (en) | Palm vein image recognition method, device, equipment and medium based on GhostNet network | |
| CN114373213A (en) | Juvenile identity recognition method and device based on face recognition | |
| CN116152938A (en) | Method, device and equipment for training identity recognition model and transferring electronic resources | |
| CN113505716A (en) | Training method of vein recognition model, and recognition method and device of vein image | |
| CN113762249A (en) | Image attack detection and image attack detection model training method and device | |
| CN115082873A (en) | Image recognition method and device based on path fusion and storage medium | |
| CN112308093A (en) | Air quality perception method based on image recognition, model training method and system | |
| CN117079336B (en) | Training method, device, equipment and storage medium for sample classification model | |
| CN114582006B (en) | Child age-crossing face recognition method and device, electronic equipment and medium | |
| CN114360008B (en) | Face authentication model generation method, authentication method, equipment and storage medium | |
| CN110245302B (en) | Policy generation method and device for identifying fraudulent cases and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |