Credibility determination method, device, equipment and storage medium
Through the integration of multi-dimensional data characteristics and unsupervised learning of financial users, the problem of poor adaptability of existing financial anti-fraud measures is solved, and the accurate mining and credibility determination of fraud is achieved, which improves the efficiency and accuracy of financial business sex.
Patent Information
- Application Number
- CN202010046908.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-01-16
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2040-01-16
AI Technical Summary
Existing financial anti-fraud methods rely on expert experience and supervised models, are difficult to adapt to the rapid changes in fraud patterns and new fraudulent behaviors, and require a large number of fraud labels to be trained, resulting in insufficient adaptability and flexibility.
By fusion of multi-dimensional data features of the object to be tested, the feature dimensions are reduced, the feature cross-comparison efficiency is improved, unsupervised learning is achieved, user credibility is determined, and dependence on known fraud labels is avoided.
It realizes accurate exploration of user fraud behavior, improves the efficiency and accuracy of credibility determination, enhances the visualization of behavioral characteristics, reduces manual verification costs, and improves the efficiency of financial business review.
Smart Images

Figure CN111222981B_ABST
Abstract
Description
Technical Field
[0001] Embodiments of the present invention relate to the field of information processing technology, and in particular to a credibility determination method, apparatus, device and storage medium. Background Art
[0002] At present, driven by emerging cutting-edge technologies such as big data, artificial intelligence, and blockchain, the penetration of financial technology into the traditional financial industry has become an irreversible trend. At the same time, the risk of financial fraud is also expanding, and the anti-fraud situation is severe. For example, before issuing a credit card to a user, the user's credit level needs to be approved to uncover credit card fraud.
[0003] At present, the commonly used financial anti-fraud methods are to use expert experience guidance or supervised anti-fraud models for determination. Expert experience guidance is based on predetermined expert rules to score risky behavior data, thereby judging risky financial behaviors. The use of supervised anti-fraud models is to obtain a prediction model through supervised training based on the behavior of predetermined users with fraud labels, so as to use the model to predict the user's financial behavior.
[0004] Expert experience is obtained by summarizing past fraud cases, but today’s fraudsters can quickly change their fraud patterns. New types of fraud are often completed by many complex events and steps, and it is difficult to quickly extract rules. Therefore, using experience to judge fraud is difficult to adapt to the evolving fraud patterns, and has poor adaptability and flexibility. Training supervised anti-fraud models requires a large number of fraud labels, and the difficulty in defining and obtaining fraud labels brings certain difficulties to the training of anti-fraud models. Summary of the invention
[0005] The embodiments of the present invention provide a credibility determination method, apparatus, device and storage medium, which achieve accurate mining of user fraudulent behavior by fusing multi-dimensional data features of the object to be tested, and do not require supervised training with known fraud labels, thereby improving the efficiency of credibility determination.
[0006] In a first aspect, an embodiment of the present invention provides a credibility determination method, comprising:
[0007] Obtaining at least two types of characteristic information of the object to be tested;
[0008] fusing the at least two types of feature information to obtain fused feature information;
[0009] Determining a feature representation of the object to be measured according to the fused feature information;
[0010] The credibility of the object to be tested is determined according to the characteristic representation.
[0011] In a second aspect, an embodiment of the present invention further provides a credibility determination device, including:
[0012] A feature information acquisition module, used to acquire at least two types of feature information of the object to be tested;
[0013] A feature information fusion module, used for fusing the at least two feature information to obtain fused feature information;
[0014] A feature representation determination module, used to determine the feature representation of the object to be measured according to the fused feature information;
[0015] The credibility determination module is used to determine the credibility of the object to be tested according to the feature representation.
[0016] In a third aspect, an embodiment of the present invention further provides a device, including:
[0017] one or more processors;
[0018] a storage device for storing one or more programs,
[0019] When the one or more programs are executed by the one or more processors, the one or more processors implement the credibility determination method as described in any embodiment of the present invention.
[0020] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the credibility determination method as described in any embodiment of the present invention.
[0021] The embodiment of the present invention is based on the fusion of multiple acquired feature information to achieve the simplification of feature information and improve the efficiency of feature cross-comparison; the fused features are represented to enhance the visualization effect of user behavior features; and the credibility of the object to be tested is determined. By fusing the multi-dimensional feature data of the object to be tested, accurate mining of user fraud behavior is achieved, and no known fraud labels are required for supervised training, thereby improving the efficiency of credibility determination. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] Figure 1 is a flow chart of a method for determining credibility in Embodiment 1 of the present invention;
[0023] Figure 2 is a flow chart of a method for determining credibility in Embodiment 2 of the present invention;
[0024] Figure 3 is a schematic diagram of the structure of a credibility determination device in Embodiment 3 of the present invention;
[0025] Figure 4It is a schematic diagram of the structure of the device in the fourth embodiment of the present invention. DETAILED DESCRIPTION
[0026] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.
[0027] Figure 1 is a flow chart of the credibility determination method in the first embodiment of the present invention. This embodiment is applicable to the case where the credibility is determined by unsupervised learning of the fused feature information of the object to be tested. The method can be executed by a credibility determination device, which can be implemented in software and / or hardware and can be configured in a device, for example, a background server or other device with communication and computing capabilities. Figure 1 As shown, the method specifically includes:
[0028] Step 101: Obtain at least two types of feature information of the object to be tested.
[0029] The object to be tested refers to the user who submits a credit event to the bank or the credit event submitted by the user to the bank system. The characteristic information of the user who submits a credit event to the bank refers to the original characteristic data obtained based on the user himself and the event.
[0030] In this embodiment, optionally, the at least two types of characteristic information include at least one of the following categories: credit operation data, credit equipment data used in the credit operation, and basic attribute information;
[0031] The credit operation data includes input data in the credit interface and / or the time spent on the credit interface;
[0032] The credit device data includes the operating system type and / or network type of the device;
[0033] The basic attribute information includes at least one of the following: education background, place of origin, marital status, bank statements, tax records and provident fund records.
[0034] Among them, the input data in the credit interface in the credit operation data refers to the input content and clicked web page content on the web page interface when the user performs a credit operation, for example, the user clicks on the product type data of a certain credit product, or the user inputs content in the search box; the credit interface residence time refers to the user's residence time on the target interface, reflecting the user's interest in the target, for example, the user's visit time on the product interface after clicking on a certain credit product; obtaining the user's credit operation data can understand the user's credit tendency.
[0035] The credit device used in the credit operation refers to the device used by the user to perform the credit operation, such as the user using a personal mobile phone to perform the credit operation or using a bank device to perform the credit operation; the operating system type of the device refers to the system type used by the device, such as Windows, macOS, Linux, iOS or Android, etc.; the network type refers to the operator to which the data used by the device belongs, such as China Mobile, China Telecom or China Unicom, etc. Obtaining the credit device data used in the credit operation can help understand the user's operating environment and increase the richness of feature information.
[0036] The basic attribute information includes the basic attributes of the subject to be tested as a natural person and its retention data in the bank. Obtaining the basic attributes of the subject to be tested can understand the demographic learning information of the user who submitted the credit event, and have a preliminary understanding of the user's assets, providing reference data for the credibility of its credit event.
[0037] Specifically, after determining the identity of the user submitting the credit operation, relevant feature information is obtained from the bank system, and other basic attribute data of the user is obtained from the corresponding official system, and the initial data is obtained as the feature information of the object to be tested. Since the obtained feature information includes the feature data of each dimension of the user, the complexity of the user's behavior can be represented, which is convenient for judging the credibility of the user's credit operation based on the user's comprehensive behavior. In order to describe the user's complex behavior more accurately, when obtaining the user's feature information, the relevant information of the user's credit behavior should be collected as comprehensively and detailed as possible, and the feature dimensions of the original features should be enriched to improve the richness of the feature information, thereby improving the accuracy of the description of the user's comprehensive behavior.
[0038] Step 102: Fuse the at least two types of feature information to obtain fused feature information.
[0039] Specifically, fusion refers to merging all acquired feature information to reduce feature dimensions. Exemplarily, based on the association relationship between the preset feature information, the associated feature information is merged according to preset rules, such as taking the average, sum, and difference. After the fusion operation, the similarity between the feature information is low, and all acquired feature information of the object to be tested is reflected with the least feature information, thereby improving the efficiency of credibility determination.
[0040] In this embodiment, optionally, fusing the at least two types of feature information to obtain fused feature information includes:
[0041] Determine the similarity between different feature information;
[0042] If the similarity between any set of feature information is greater than the similarity threshold, the set of feature information is fused to obtain fused feature information.
[0043] Among them, similarity refers to the specific contribution direction of different feature information in measuring the credibility of the object to be tested. For example, the salary income and provident fund information of users who submit credit operations have a certain relationship, and the information they reflect is essentially the same.
[0044] Specifically, the similarity between different feature information can be determined based on the correlation coefficient between the feature information, or the contribution of different feature information to the credibility determination can be determined based on variance analysis. If it is determined that the similarity between any set of feature information is large, the set of information is merged, for example, a set of feature information is merged into a comprehensive feature information to reduce the amount of feature information.
[0045] Optionally, the amount of fused feature information is less than the amount of initial feature information. The initial feature information refers to the feature information directly obtained from the object to be tested. After fusion, the amount of feature information is reduced, that is, the feature data is subjected to dimensionality reduction processing, which improves the efficiency of similarity calculation and data visualization in cross-matching of behavioral features. In this embodiment, the specific operation of fusion is not limited, and certain merging operations on the acquired feature information by certain means are all within the protection scope of the present invention.
[0046] Step 103: Determine the feature representation of the object to be measured according to the fused feature information.
[0047] Specifically, after the initial feature information is fused, the amount of feature information is reduced, and fused feature information is obtained, and the fused feature information is relatedly represented to facilitate subsequent credibility determination operations. Exemplarily, the fused feature information is represented by a feature vector to accurately reflect the fused feature information of each dimension. For example, the fused feature information is determined as a feature vector of the object to be tested according to a certain mapping relationship, and the size of the feature vector is preset. The feature vector can be determined by word embedding.
[0048] The feature representation is determined based on the fused feature information, so that the feature representation of the object to be tested can more accurately describe the user's behavior sequence characteristics. Since the determination of the credibility of the user's credit behavior is equivalent to predicting the user's behavior, improving the accuracy of the description of the user's behavior can improve the accuracy of the credibility determination.
[0049] Step 104: Determine the credibility of the object to be tested according to the characteristic representation.
[0050] Specifically, the credibility of the object to be tested is determined based on the description of the behavior of the object to be tested in the feature representation. Exemplarily, the behavioral feature representation of a user with high credibility is determined in advance, and the similarity between the feature representation of the object to be tested and the similarity is determined. If the similarity is greater than a preset threshold, it means that the credibility of the object to be tested is credible; on the contrary, if the similarity is less than the preset threshold, it means that the credibility of the object to be tested is unreliable. Determining the feature representation of a user with known high credibility reflects the behavioral characteristics of a normal user. Based on it, it is determined whether the behavior of the object to be tested deviates from the behavioral characteristics of a normal user. If the deviation is large, it means that the behavior of the object to be tested is abnormal and the credibility is not high. By analyzing the feature representation of the object to be tested, the potential behavioral characteristics of the object to be tested can be mined to improve the accuracy of credibility determination.
[0051] The embodiment of the present invention is based on the fusion of multiple acquired feature information to achieve the simplification of feature information and improve the efficiency of feature cross-comparison; the fused features are represented to enhance the visualization effect of user behavior features; and the credibility of the object to be tested is determined. By fusing the multi-dimensional feature data of the object to be tested, accurate mining of user fraud behavior is achieved, and no known fraud labels are required for supervised training, thereby improving the efficiency of credibility determination.
[0052] Embodiment 2
[0053] Figure 2 is a flow chart of a method for determining credibility in Embodiment 2 of the present invention. Embodiment 2 is further optimized on the basis of Embodiment 1. Figure 2 As shown, the method includes:
[0054] Step 201: Obtain at least two types of characteristic information of the object to be tested.
[0055] Step 202: Fuse the at least two types of feature information to obtain fused feature information.
[0056] Step 203: Determine the feature representation of the object to be measured according to the fused feature information.
[0057] Specifically, embedding learning is performed on the feature information of the object to be tested, and pre-training is performed to determine the embedding vector of the object to be tested. The feature representation obtained through embedding learning can more accurately express the characteristics of the behavior and environment of the object to be tested. Exemplarily, the word embedding idea is used to pre-train the user's feature information through an autoencoder network to generate a behavior embedding model, thereby determining the embedding vector obtained according to the feature information of the object to be tested.
[0058] Exemplarily, the feature information of the object to be tested is pre-trained through the encoding part of the autoencoder network to generate an embedding vector. Through deep learning, a more accurate feature expression of the object to be tested is output, and the acquired high-dimensional data is reduced in dimension, which improves the efficiency of similarity calculation and data visualization in cross-matching of behavioral features. For example, the embedding vector pre-training part uses an autoencoder network containing two hidden layers to train the behavioral embedding model, and the generation of the embedding vector uses the encoding network part, which contains an input layer and two fully connected hidden layers.
[0059] The embedding vector generated through pre-training reflects the behavioral portrait of the object to be tested, builds user portraits more efficiently, deeply mines features such as user behavior and environmental information, and improves the accuracy of credibility determination.
[0060] Step 204: Based on the anomaly detection model, determine the credibility probability of the object to be predicted according to the feature representation.
[0061] The anomaly detection model is obtained by using positive sample objects to perform unsupervised training on a neural network structure.
[0062] Positive sample objects refer to users with high credibility. The feature representations generated by the feature information of these users are determined, and the feature representations of the positive sample objects are used for training. Since it is difficult to define a clear definition of a user's fraud label and it takes a long time to determine, it is difficult to obtain users with low credibility. In this embodiment, only positive sample objects need to be used for training to improve the efficiency of model determination.
[0063] Specifically, the anomaly detection model uses a two-layer autoencoder network to perform unsupervised training on the embedding vectors of the positive sample objects. The network learns the behavioral characteristics of the positive sample objects and uses this as a basis to judge the deviation of the behavioral characteristics of the object to be tested.
[0064] The autoencoder network consists of two parts: the encoding recognition network and the decoding generation network. The encoding part maps the input vector to the latent space through the encoding function, while the decoding part reconstructs the latent space to the input space through the decoding function. The overall optimization goal is to minimize the mean square error of compression and reconstruction.
[0065] The encoding part uses the hypothesis function of the standard neural network, that is, the nonlinear mapping of the linear function, and the hypothesis function of the decoding with a symmetric structure is similar. The optimization goal of the network is to make the output after encoding and decoding as close to the original input space as possible. The weights are continuously updated iteratively by forward transmission and backward feedback until the loss function converges.
[0066] The autoencoder network is a multi-layer feedforward neural network. Its input and output are the same, and no data labels are required, so as to achieve the effect of unsupervised training. It consists of two parts: encoding part and decoding part. Among them, the encoding part is also a cognitive network, which reduces the dimension of the input data layer by layer, which is equivalent to an encoding process of the original data, forming a condensed representation of the original data. The decoding part is also called the generative network. After reaching the lowest dimension, neurons are added layer by layer, which is equivalent to a decoding process. The decoding output should be as close to the original data as possible, that is, under the premise of greatly compressing the original feature space, the compression loss should be minimized. The purpose of this network is to understand the data more clearly. Under the guidance of this optimization goal, the learned network connection weights are all different embedding vectors. Through deep learning, how to better express the characteristics of the data, the autoencoder network has the effect of dimensionality reduction in the middle hidden layer, so it forces the network to learn the internal pattern of the data, and plays an efficient role in memorizing the training data. The symmetric network structure also allows weights to be shared in encoding and decoding training, greatly improving the training speed.
[0067] Exemplarily, after the anomaly detection model is generated through the embedding vector training of the positive sample object, the embedding vector of the object to be tested is input into the anomaly detection model, and the model can detect outliers that deviate greatly from the behavioral characteristics learned by the network, output the credibility probability, and realize the anomaly detection of the object to be tested. Since the anomaly detection model reflects the normal behavioral characteristics of the positive sample object, it can, to a certain extent, mine new fraud behaviors and fraud patterns, improve the accuracy of determining the credibility of the object to be tested, and reduce the loss of the bank.
[0068] Step 205: Determine the credibility of the object to be tested according to the credibility probability.
[0069] Specifically, the credibility of the object to be tested is determined according to a predetermined mapping relationship between credibility probability and credibility. Exemplarily, when the credibility probability is greater than 0.8, the credibility of the object to be tested is determined to be credible, and when it is less than or equal to 0.8, the credibility of the object to be tested is determined to be untrustworthy. Further subdivision can also be performed, for example, when the credibility probability is greater than 0.8, the credibility of the object to be tested is determined to be credible, when it is greater than 0.5 and less than or equal to 0.8, the credibility of the object to be tested is determined to be uncertain, and when it is less than or equal to 0.5, the credibility of the object to be tested is determined to be untrustworthy.
[0070] The embodiment of the present invention generates a user behavior profile by pre-training the feature information of the object to be tested, and then performs unsupervised training on the embedded vector in the user behavior profile through a multi-layer autoencoder network to predict the credibility probability. By using the embedded vector representation of the multi-dimensional feature data of the object to be tested, accurate mining of user fraud behavior is achieved, and no known fraud labels are required for supervised training, thereby improving the efficiency of credibility determination. In addition, the credibility of the object to be tested is determined by the predicted credibility probability, which reduces the labor cost of subsequent verification and review. At the same time, it can also assist in verifying the credit score of the object to be tested, thereby improving the efficiency of banking business review.
[0071] Embodiment 3
[0072] Figure 3 is a schematic diagram of the structure of the credibility determination device in the third embodiment of the present invention. This embodiment is applicable to the case where the credibility is determined by performing unsupervised learning on the fused feature information of the object to be tested. Figure 3 As shown, the device comprises:
[0073] The feature information acquisition module 310 is used to acquire at least two types of feature information of the object to be tested;
[0074] A feature information fusion module 320, configured to fuse the at least two types of feature information to obtain fused feature information;
[0075] A feature representation determination module 330, used to determine the feature representation of the object to be detected according to the fused feature information;
[0076] The credibility determination module 340 is used to determine the credibility of the object to be tested according to the feature representation.
[0077] The embodiment of the present invention is based on the fusion of multiple acquired feature information to achieve the simplification of feature information and improve the efficiency of feature cross-comparison; the fused features are represented to enhance the visualization effect of user behavior features; and the credibility of the object to be tested is determined. By fusing the multi-dimensional feature data of the object to be tested, accurate mining of user fraud behavior is achieved, and no known fraud labels are required for supervised training, thereby improving the efficiency of credibility determination.
[0078] Optionally, the at least two types of characteristic information include at least one of the following categories: credit operation data, credit equipment data used in the credit operation, and basic attribute information;
[0079] The credit operation data includes input data in the credit interface and / or the time spent on the credit interface;
[0080] The credit device data includes the operating system type and / or network type of the device;
[0081] The basic attribute information includes at least one of the following: education background, place of origin, marital status, bank statements, tax records and provident fund records.
[0082] Optionally, the feature information fusion module 320 is specifically used for:
[0083] Determine the similarity between different feature information;
[0084] If the similarity between any set of feature information is greater than the similarity threshold, the set of feature information is fused to obtain fused feature information.
[0085] Optionally, the credibility determination module 340 is specifically configured to:
[0086] Based on the anomaly detection model, determining the credibility probability of the object to be predicted according to the feature representation;
[0087] Determining the credibility of the object to be tested according to the credibility probability;
[0088] The anomaly detection model is obtained by using positive sample objects to perform unsupervised training on a neural network structure.
[0089] The credibility determination device provided in the embodiment of the present invention can execute the credibility determination method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the credibility determination method.
[0090] Embodiment 4
[0091] Figure 4 It is a structural schematic diagram of a device provided in Embodiment 4 of the present invention. Figure 4 A block diagram of an exemplary device 12 suitable for use in implementing embodiments of the present invention is shown. Figure 4 The device 12 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.
[0092] like Figure 4 As shown, device 12 is in the form of a general purpose computing device. Components of device 12 may include, but are not limited to, one or more processors or processing units 16, system memory 28, and a bus 18 that connects various system components (including system memory 28 and processing units 16).
[0093] Bus 18 represents one or more of several types of bus structures, including a storage device bus or storage device controller, a peripheral bus, an accelerated graphics port, a processor or a local bus using any of a variety of bus architectures. By way of example, these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.
[0094] Device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by device 12, including volatile and non-volatile media, removable and non-removable media.
[0095] System storage 28 may include computer system readable media in the form of volatile storage devices, such as random access memory (RAM) 30 and / or cache memory 32. Device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Figure 4 not shown, usually called a "hard drive"). Although Figure 4 Not shown in the figure, a disk drive for reading and writing to a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, a DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The storage device 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present invention.
[0096] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in the storage device 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described herein.
[0097] The device 12 may also communicate with one or more external devices 14 (e.g., keyboards, pointing devices, displays 24, etc.), may communicate with one or more devices that enable a user to interact with the device 12, and / or may communicate with any device that enables the device 12 to communicate with one or more other computing devices (e.g., network cards, modems, etc.). Such communication may be performed via an input / output (I / O) interface 22. Furthermore, the device 12 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with other modules of the device 12 via a bus 18. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the device 12, including, but not limited to, microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.
[0098] The processing unit 16 executes various functional applications and data processing by running the programs stored in the system storage device 28, for example, implementing the credibility determination method provided in the embodiment of the present invention, including:
[0099] Obtaining at least two types of characteristic information of the object to be tested;
[0100] fusing the at least two types of feature information to obtain fused feature information;
[0101] Determining a feature representation of the object to be measured according to the fused feature information;
[0102] The credibility of the object to be tested is determined according to the characteristic representation.
[0103] Embodiment 5
[0104] Embodiment 5 of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the method for determining the credibility provided in the embodiment of the present invention is implemented, including:
[0105] Obtaining at least two types of characteristic information of the object to be tested;
[0106] fusing the at least two types of feature information to obtain fused feature information;
[0107] Determining a feature representation of the object to be measured according to the fused feature information;
[0108] The credibility of the object to be tested is determined according to the characteristic representation.
[0109] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.
[0110] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.
[0111] Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
[0112] Computer program code for performing the operations of the present invention may be written in one or more programming languages or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).
[0113] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims
1. A credibility determination method, characterized in that: include: Obtaining at least two types of characteristic information of the object to be tested; fusing the at least two types of feature information to obtain fused feature information; Determine the feature representation of the object to be tested according to the fused feature information, pre-train the feature information of the object to be tested through the encoding part of the autoencoder network, and generate an embedding vector of the object to be tested; Determining the credibility of the object to be tested according to the characteristic representation; Determining the credibility of the object to be tested according to the characteristic representation includes: Based on the anomaly detection model, determining the credibility probability of the object to be detected according to the feature representation; The credibility of the object to be tested is determined according to the credibility probability; wherein the credibility probability is obtained by generating an anomaly detection model through the embedding vector of the positive sample object, inputting the embedding vector of the object to be tested into the anomaly detection model, and the model detects outliers that deviate greatly from the behavioral characteristics learned by the network; wherein the positive sample object refers to a user with a determined high credibility; The anomaly detection model uses a two-layer autoencoder network to perform unsupervised training on the embedding vector of the positive sample object. The network learns the behavioral characteristics of the positive sample object and uses this as a basis to judge the deviation of the behavioral characteristics of the object to be tested. The at least two types of characteristic information include at least one of the following: credit operation data, credit equipment data used in the credit operation, and basic attribute information; wherein the basic attribute information includes the basic attributes of the subject to be tested as a natural person and its retained data in the bank; The credit operation data includes input data in the credit interface and / or the time spent on the credit interface; The credit device data includes the operating system type and / or network type of the device; The basic attribute information includes at least one of the following: education background, place of origin, marital status, bank statements, tax records and provident fund records.
2. The method according to claim 1, characterized in that The at least two types of feature information are fused to obtain fused feature information, including: Determine the similarity between different feature information; If the similarity between any set of feature information is greater than the similarity threshold, the set of feature information is fused to obtain fused feature information.
3. A credibility determination device, characterized in that: include: A feature information acquisition module, used to acquire at least two types of feature information of the object to be tested; A feature information fusion module, used for fusing the at least two feature information to obtain fused feature information; A feature representation determination module is used to determine the feature representation of the object to be tested according to the fused feature information, pre-train the feature information of the object to be tested through the encoding part of the autoencoder network, and generate an embedding vector of the object to be tested; A credibility determination module, used to determine the credibility of the object to be tested according to the feature representation; The credibility determination module is specifically used for: Based on the anomaly detection model, determining the credibility probability of the object to be detected according to the feature representation; The credibility of the object to be tested is determined according to the credibility probability; wherein the credibility probability is obtained by generating an anomaly detection model through the embedding vector of the positive sample object, inputting the embedding vector of the object to be tested into the anomaly detection model, and the model detects outliers that deviate greatly from the behavioral characteristics learned by the network; wherein the positive sample object refers to a user with a determined high credibility; The anomaly detection model uses a two-layer autoencoder network to perform unsupervised training on the embedding vector of the positive sample object. The network learns the behavioral characteristics of the positive sample object and uses this as a basis to judge the deviation of the behavioral characteristics of the object to be tested. The at least two types of characteristic information include at least one of the following: credit operation data, credit equipment data used in the credit operation, and basic attribute information; wherein the basic attribute information includes the basic attributes of the subject to be tested as a natural person and its retained data in the bank; The credit operation data includes input data in the credit interface and / or the time spent on the credit interface; The credit device data includes the operating system type and / or network type of the device; The basic attribute information includes at least one of the following: education background, place of origin, marital status, bank statements, tax records and provident fund records.
4. The device according to claim 3, characterized in that Feature information fusion module, specifically used for: Determine the similarity between different feature information; If the similarity between any set of feature information is greater than the similarity threshold, the set of feature information is fused to obtain fused feature information.
5. A device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the credibility determination method as described in any one of claims 1-2.
6. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the credibility determination method as described in any one of claims 1-2 is implemented.
