WO2020211388A1

WO2020211388A1 - Behavior prediction method and device employing prediction model, apparatus, and storage medium

Info

Publication number: WO2020211388A1
Application number: PCT/CN2019/121771
Authority: WO
Inventors: 胡艺飞; 徐国强; 邱寒
Original assignee: 深圳壹账通智能科技有限公司
Priority date: 2019-04-16
Filing date: 2019-11-28
Publication date: 2020-10-22
Also published as: CN110222554A

Abstract

A behavior prediction method and device employing a prediction model, an electronic apparatus, and a storage medium. The method comprises: upon receiving a fraud identification instruction, obtaining at least one video to be reviewed (S10); extracting a target video sequence for each of the at least one video to be reviewed (S11); inputting each target video sequence into a pre-trained prediction model, and determining the probability that a target person is committing fraud in each target video sequence (S12); combining, on the basis of a quantile principle, the respective probabilities into a target vector (S13); inputting the target vector into a pre-trained classifier, and determining a target probability (S14); and determining, according to the target probability, whether or not the target person represents a fraud risk (S15).

Description

Behavior prediction method, device, equipment and storage medium based on prediction model

This application claims the priority of the Chinese patent application filed with the Chinese Patent Office on April 16, 2019 with the application number 201910306022.3 and the invention titled "Fraud Recognition Method, Device, Electronic Equipment and Storage Medium", the entire content of which is incorporated by reference In this application.

Technical field

This application relates to the technical field of intelligent decision-making, and in particular to a behavior prediction method, device, electronic device, and storage medium based on a prediction model.

Background technique

In the prior art solution, when judging whether a customer is at risk of fraud, it is usually only identified by combining the emotions displayed by the customer’s facial expressions, and only one problem is judged. Since the judgment basis is relatively single, the reliability of the judgment result is Lower. Therefore, the user cannot directly use the above judgment result as a conclusion whether the customer has fraudulent behavior, which causes inconvenience to the user and is not conducive to the user experience.

Summary of the invention

In view of the above, it is necessary to provide a behavior prediction method, device, electronic device, and storage medium based on a predictive model. This application can make a judgment based on multiple questions and determine whether the target person is at risk of fraud, thereby realizing fraudulent behavior Due to the diversification of data, the judgment is more accurate and credible, which brings a better experience to users.

In the first aspect, this application provides a behavior prediction method based on a prediction model, the method including:

When receiving a fraud identification instruction, obtain at least one video to be detected;

Extracting the target video sequence of each video to be detected in the at least one video to be detected;

Input each target video sequence into a pre-trained prediction model to determine the fraud probability of the target person in each target video sequence;

Based on the quantile principle, combine each fraud probability into a target vector;

Input the target vector into a pre-trained classifier to determine the target probability;

According to the target probability, it is determined whether the target person has a risk of fraud.

In a second aspect, the present application also provides a behavior prediction device based on a prediction model, the device including:

The obtaining unit is configured to obtain at least one video to be detected when a fraud identification instruction is received;

An extraction unit, configured to extract the target video sequence of each video to be detected in the at least one video to be detected;

The determining unit is used to input each target video sequence into a pre-trained prediction model to determine the fraud probability of the target person in each target video sequence;

The combination unit is used to combine each fraud probability into a target vector based on the quantile principle;

The determining unit is further configured to input the target vector into a pre-trained classifier to determine the target probability;

The determining unit is further configured to determine whether the target person has a risk of fraud according to the target probability.

In a third aspect, the present application also provides an electronic device that includes a memory and a processor; the memory is used to store computer-readable instructions; the processor is used to execute the computer program and execute The computer program implements the behavior prediction method based on the prediction model as described above.

In a fourth aspect, the present application also provides a computer-readable storage medium having at least one instruction stored in the computer-readable storage medium, and the at least one instruction is executed by a processor in an electronic device to implement the aforementioned The behavior prediction method based on the prediction model.

It can be seen from the above technical solutions that this application can obtain at least one video to be detected when a fraud identification instruction is received, so as to make a judgment in combination with multiple questions, and further extract each of the at least one video to be detected. In order to remove redundant information and avoid the burden of redundant information on the operation of the electronic device, each target video sequence is input into a pre-trained prediction model, and the target person is determined in each target video sequence. Based on the quantile principle, each fraud probability is combined into a target vector, and the target vector is further input into a pre-trained classifier to determine the target probability, and according to the target probability, determine the target probability. Whether the target person is at risk of fraud, so as to realize automatic judgment of fraudulent behavior, and because of the diversification of data, the judgment is more accurate and credible, which brings a better experience to users.

Description of the drawings

Fig. 1 is a flowchart of a preferred embodiment of a behavior prediction method based on a prediction model of the present application.

Fig. 2 is a functional block diagram of a preferred embodiment of a behavior prediction device based on a prediction model of the present application.

FIG. 3 is a schematic structural diagram of an electronic device implementing a preferred embodiment of a method for predicting behavior based on a predictive model in this application.

detailed description

In order to make the objectives, technical solutions, and advantages of the present application clearer, the present application will be described in detail below with reference to the accompanying drawings and specific embodiments.

As shown in FIG. 1, it is a flowchart of a preferred embodiment of the behavior prediction method based on the prediction model of the present application. According to different needs, the order of the steps in the flowchart can be changed, and some steps can be omitted.

The behavior prediction method based on the prediction model is applied to one or more electronic devices, and the method includes:

S10: Obtain at least one video to be detected when a fraud identification instruction is received.

In the prior art solution, when judging whether a customer is at risk of fraud, it is usually only identified by combining the emotions displayed by the customer’s facial expressions, and only one problem is judged. Since the judgment basis is relatively single, the reliability of the judgment result is Lower. Therefore, the user cannot directly use the above judgment result as a conclusion whether the customer has fraudulent behavior, which causes inconvenience to the user.

In at least one embodiment of the present application, receiving the fraud identification instruction by the electronic device includes, but is not limited to, one or a combination of the following:

(1) When the electronic device detects that a customer triggers the interview process, it is determined that the electronic device receives the fraud identification instruction.

For example, when a bank conducts a client loan review, it needs to conduct a face-to-face review of the client who wants to borrow, mainly through video questioning. Specifically, the client answers the questions on the display of the electronic device one by one. From this, determine whether the customer is at risk of fraud. Therefore, when the electronic device detects that a customer is undergoing an interview, it can immediately trigger the fraud identification instruction.

Through the foregoing implementation manners, the electronic device can automatically respond when a customer conducts a face-to-face audit, avoiding human operations, and responding in a timely manner to avoid missed inspections.

(2) When the electronic device receives the signal that the configuration button is triggered, it is determined that the electronic device has received the fraud identification instruction.

Specifically, the configuration button refers to a pre-configured trigger button, and the configuration button is used to trigger the fraud identification instruction.

Further, the configuration button may be a physical button or a virtual button, which is determined according to the actual situation of the electronic device, which is not limited in this application.

Through the foregoing implementation manners, the electronic device can trigger the fraud identification instruction according to the user's requirements to meet the actual needs of the user. The conditional trigger mode also saves the operating memory of the electronic device and further improves the electronic device. Equipment performance.

In at least one embodiment of the present application, the at least one video to be detected includes, but is not limited to, one or a combination of the following:

(1) In the process of bank business face-to-face review, the answer video for each question.

It is understandable that for banks, many businesses have certain risks. For example, when a customer makes a loan, the bank needs to judge whether the customer has a credit problem in the past, so as to determine whether the customer who wants the loan is credible and repay it on time. How likely is it to make a loan, etc., and further determine whether the loan should be given to the customer.

Therefore, in view of the above situation, the bank will conduct face-to-face audits of customers with similar needs. The face-to-face review process includes several questions for customers to answer.

In the customer answering process, the electronic device can record the customer's answering each question as the at least one video to be detected.

(2) During the face-to-face review of the insurance business, the answer video for each question.

It is understandable that for insurance companies, many insurance businesses also need to review the customer's creditworthiness (whether there has been an overdue behavior, etc.) to ensure the insurance company's own interests. For example, when a customer performs insurance claim settlement services, in order to avoid fraudulent insurance, the insurance company needs to verify the customer's creditworthiness to determine whether the customer is credible and further determine whether to settle the claim.

Therefore, similar to the situation in (1) above, an insurance company can also conduct a face-to-face review of customers with similar needs. The face-to-face review process includes several questions for customers to answer.

In addition, during the customer's answering process, the electronic device records the customer's answering each question as the at least one to-be-detected video.

It should be noted that in other embodiments, the video to be detected may also include other types or fields, which is not limited in this application.

S11: Extract a target video sequence of each video to be detected in the at least one video to be detected.

In at least one embodiment of the present application, since the length of the video to be detected is different and contains a lot of useless information, the electronic device must extract the video to be detected to remove redundant information and avoid redundant information. The information puts a burden on the operation of the electronic device, affects the operation speed of the electronic device, and further affects the execution of the fraud identification instruction.

Preferably, the electronic device extracts the at least one video to be detected, and the target video sequence of each video to be detected includes:

The electronic device uses a K-means clustering algorithm (K-means clustering algorithm) to extract the target video sequence of each of the at least one video to be detected.

Specifically, the electronic device adopts the K-means clustering algorithm, can extract the key video sequence in the video to be detected, and can ensure that the extracted feature dimensions of each video to be detected are consistent, thereby satisfying Algorithm requirements.

Of course, in other embodiments, as long as the same video extraction effect can be achieved, the electronic device may also use other algorithms, and this application is not limited.

Through the foregoing implementation manner, the electronic device can extract the target video sequence from the video to be detected for use in subsequent processes.

S12: Input each target video sequence into a pre-trained prediction model, and determine the fraud probability of the target person in each target video sequence.

In at least one embodiment of the present application, the prediction model is a prediction model integrating multiple attributes, and the prediction model is pre-trained by the electronic device.

Specifically, before the electronic device inputs each target video sequence into a pre-trained prediction model to determine the fraud probability of the target person in each target video sequence, the method further includes:

The electronic device obtains a sample video sequence. Further, the electronic device uses a face recognition algorithm to extract the facial movements and eyeball angles of each person in the sample video sequence. Furthermore, the electronic device uses a support vector The regression algorithm trains the facial actions of each character to obtain the first vector of each character, and uses the neural network algorithm to train the eyeball angles of all the characters to obtain the second vector of each character, and the electronic device intercepts all the characters. The head rotation angle of the character is obtained, and the third vector of each character is obtained, and the first vector, second vector, and third vector of each character are combined to obtain the sample vector of each character. The electronic device obtains the sample vector from the configuration database. , Determine the overdue data of each character to formulate the overdue label of each character, and further, the electronic device uses the sample vector and overdue label of each character as sample data, and uses gradient boosting to train the Forecast model.

Wherein, the sample video sequence may include any data with human faces acquired by the electronic device, and may also include customer data stored in a designated database (such as a bank's database, an insurance company's database, etc.), as long as it has Sufficient sample size, this application is not limited.

Preferably, extracting the eyeball angle of each person in the sample video sequence by the electronic device includes:

The electronic device scales all pictures in the target video sequence to the same size, and removes pictures whose eyes cannot be seen due to excessive head rotation or light. The electronic device uses face feature point detection technology to intercept pictures of eye parts to obtain each face eye picture, and record the polar coordinate value of the head rotation and what the eye sees in each face eye picture Direction to get the eyeball angle.

Further, the electronic device may use a support vector regression algorithm (linear-svr) to train 15 motion features of the human face, such as the intensity of blinking, frowning, etc., and control the intensity within the range of 0-5, and the value The higher the higher the intensity. The electronic device further uses the k-means clustering algorithm to extract 5 target video sequences from the video to be detected to obtain 5*15 key features, which are spliced together in chronological order to obtain a 75-length first vector .

Furthermore, the electronic device uses a neural network algorithm to train the eye angles of all the characters. Specifically, the electronic device inputs the eye pictures in the training set to the convolutional neural network model for training, specifically using three consecutive layers The convolution and maximum pooling layer is connected to a fully connected layer at the end, where the polar coordinates of the head rotation are spliced on the fully connected layer of the last layer. Finally, the degree of the up and down, left and right deviation of the eyeball is used as the final output, and the output result of the eyeball relative to the eye position is obtained. The network parameters of the model are adjusted according to the difference between the eyeball relative to the eye position result and the pre-marked eyeball relative to the eye position. The network parameters of the model are continuously optimized to determine the neural network model, and the electronic device inputs the eyeball angles of all the characters to the neural network model to obtain the second vector of each character.

Furthermore, after the electronic device intercepts the head rotation angles of all the characters, it sets a certain deviation angle threshold and deviation times threshold, that is, it exceeds the preset angle and reaches the preset number as 1, otherwise Mark it as 0 to get the third vector.

Furthermore, the configuration database can store the overdue data of each character. When a customer has an overdue behavior, an overdue label is configured for it, and when a customer has no overdue behavior, an overdue label is configured for the customer.

Of course, in other embodiments, the prediction model may also adopt other training methods, which is not limited in this application.

S13, based on the quantile principle, combine each fraud probability into a target vector.

In at least one embodiment of the present application, the electronic device adopts the quantile principle to concatenate each fraud probability into a vector in chronological order.

Specifically, since the classifier to be used subsequently is trained with vector values as input, the electronic device combines the target vector as the input of the classifier, so that the classifier can be used.

For example, take 20%, 40%, 60%, 80%, and 100% of each fraud probability as the feature, the length is 5, and the combination is the target vector.

S14. Input the target vector into a pre-trained classifier to determine the target probability.

In at least one embodiment of the present application, the classifier may include, but is not limited to, a linear SVM classifier (linear SVM classifier) and the like.

Specifically, the method of constructing the SVM classifier mainly includes the direct method and the indirect method, and the indirect method includes the one-to-many method and the one-to-one method.

The SVM classifier takes vectors as input.

In this embodiment, the target vector with a length of 5 is used as the input data of the SVM classifier.

S15: Determine whether the target person has a risk of fraud according to the target probability.

In at least one embodiment of the present application, the target probability may be a value between 0-1, and the higher the value, the higher the risk of fraud.

In at least one embodiment of the present application, the electronic device may be configured with a preset threshold in advance, so as to determine whether the target person is at risk of fraud by comparing with the preset threshold.

Specifically, the electronic device determining whether the target person is at risk of fraud according to the target probability includes:

When the target probability is greater than or equal to a preset threshold, the electronic device determines that the target person is at risk of fraud;

When the target probability is less than the preset threshold, the electronic device determines that the target person has no risk of fraud.

Wherein, the value of the preset threshold can be customized. For example, the preset threshold can be 0.7, which is not limited by this application.

In at least one embodiment of the present application, when it is determined that the target person is at risk of fraud, the method further includes:

The electronic device saves the video corresponding to the target person. Further, the electronic device intercepts the image of the target person from the saved video. Further, the electronic device combines the intercepted image to send a prompt message to Specify the terminal device.

Specifically, the designated terminal device may include a terminal device of a security management and control personnel, so that it can be involved in processing in time. By sending the intercepted image and prompt information to the terminal device together, it can quickly find the target person and save the video as a basis to assist the safety management personnel to make a quick judgment.

The electronic device obtains all the record information of the target person from the configuration database, and further, the electronic device sends all the record information to the designated terminal device.

Specifically, all the record information may include, but is not limited to: transaction records, credit records, etc.

Through the foregoing implementation manners, the electronic device can provide more data to assist relevant personnel in judging whether there is a risk of fraud, so that the credibility of the judgment result is higher.

As shown in FIG. 2, it is a functional block diagram of a preferred embodiment of a behavior prediction device based on a prediction model of the present application. The behavior prediction device 11 based on the prediction model includes an acquisition unit 110, an extraction unit 111, a determination unit 112, a combination unit 113, a training unit 114, an interception unit 115, a merging unit 116, a formulation unit 117, a storage unit 118, and a sending unit 119. . The module/unit referred to in this application refers to a series of computer program segments that can be executed by the processor 13 and can complete fixed functions, and are stored in the memory 12. In this embodiment, the functions of each module/unit will be described in detail in subsequent embodiments.

When a fraud identification instruction is received, the obtaining unit 110 obtains at least one video to be detected.

The extraction unit 111 extracts the target video sequence of each video to be detected in the at least one video to be detected.

Preferably, the extraction unit 111 extracts the at least one video to be detected, and the target video sequence of each video to be detected includes:

The extraction unit 111 uses a K-means clustering algorithm (K-means clustering algorithm) to extract the target video sequence of each of the at least one video to be detected.

Of course, in other embodiments, as long as the same video extraction effect can be achieved, the extraction unit 111 may also use other algorithms, and this application is not limited.

Each target video sequence is input into a pre-trained prediction model, and the determining unit 112 determines the fraud probability of the target person in each target video sequence.

In at least one embodiment of the present application, the prediction model is a prediction model integrating multiple attributes, and the prediction model is pre-trained by the training unit 114.

Specifically, before the determining unit 112 inputs each target video sequence into a pre-trained prediction model to determine the fraud probability of the target person in each target video sequence, the method further includes:

The acquisition unit 110 acquires a sample video sequence; the extraction unit 111 uses a face recognition algorithm to extract the facial movements and eyeball angles of each person in the sample video sequence; the training unit 114 uses a support vector regression algorithm to train each The facial movements of the characters obtain the first vector of each character, and use the neural network algorithm to train the eyeball angles of all the characters to obtain the second vector of each character. The intercepting unit 115 intercepts the head rotation of all the characters Angle, obtain the third vector of each character; the merging unit 116 merges the first vector, the second vector and the third vector of each character to obtain the sample vector of each character; the formulation unit 117 determines each character from the configuration database The overdue data of the characters is used to formulate an overdue label for each character; the training unit 114 uses the sample vector and overdue label of each character as sample data, and uses a gradient boosting algorithm to train the prediction model.

Based on the quantile principle, the combination unit 113 combines each fraud probability into a target vector.

According to the target probability, the determining unit 112 determines whether the target person is at risk of fraud.

Specifically, the determining unit 112 determining whether the target person is at risk of fraud according to the target probability includes:

When the target probability is greater than or equal to a preset threshold, the determining unit 112 determines that the target person is at risk of fraud;

When the target probability is less than the preset threshold, the determining unit 112 determines that the target person has no risk of fraud.

The saving unit 118 saves the video corresponding to the target person; the intercepting unit 115 intercepts the image of the target person from the saved video; the sending unit 119 combines the intercepted image to send prompt information to a designated terminal device.

The acquiring unit 110 acquires all the record information of the target person from the configuration database; the sending unit 119 sends all the record information to the designated terminal device.

As shown in FIG. 3, it is a schematic structural diagram of an electronic device in a preferred embodiment of the present application for implementing a behavior prediction method based on a prediction model.

The electronic device 1 is a device that can automatically perform numerical calculation and/or information processing according to pre-set or stored instructions. Its hardware includes, but is not limited to, a microprocessor, an Application Specific Integrated Circuit (ASIC) ), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

The electronic device 1 can also be, but is not limited to, any electronic product that can interact with the user through a keyboard, a mouse, a remote control, a touch panel, or a voice control device, for example, a personal computer, a tablet computer, or a smart phone. , Personal Digital Assistant (PDA), game consoles, interactive network TV (Internet Protocol Television, IPTV), smart wearable devices, etc.

The electronic device 1 may also be a computing device such as a desktop computer, a notebook, a palmtop computer, and a cloud server.

The network where the electronic device 1 is located includes, but is not limited to, the Internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (Virtual Private Network, VPN), etc.

In an embodiment of the present application, the electronic device 1 includes, but is not limited to, a memory 12, a processor 13, and a computer program stored in the memory 12 and running on the processor 13, such as Behavior prediction program based on prediction model.

Those skilled in the art can understand that the schematic diagram is only an example of the electronic device 1 and does not constitute a limitation on the electronic device 1. It may include more or less components than those shown in the figure, or combine certain components, or different components. Components, for example, the electronic device 1 may also include input and output devices, network access devices, buses, and the like.

The processor 13 may be a central processing unit (Central Processing Unit, CPU), other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (ASIC), Ready-made programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The processor 13 is the computing core and control center of the electronic device 1 and connects the entire electronic device with various interfaces and lines. Each part of 1, and executes the operating system of the electronic device 1, and various installed applications, program codes, etc.

The processor 13 executes the operating system of the electronic device 1 and various installed applications. The processor 13 executes the application program to implement the steps in the foregoing embodiments of the behavior prediction method based on the prediction model.

Alternatively, when the processor 13 executes the computer program, the function of each module/unit in the foregoing device embodiments is realized.

Exemplarily, the computer program may be divided into one or more modules/units, and the one or more modules/units are stored in the memory 12 and executed by the processor 13 to complete this Application. The one or more modules/units may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device 1.

The memory 12 may be used to store the computer program and/or module, and the processor 13 runs or executes the computer program and/or module stored in the memory 12 and calls the data stored in the memory 12, Various functions of the electronic device 1 are realized. The memory 12 may mainly include a storage program area and a storage data area. The storage program area may store an operating system, an application program required by at least one function (such as a sound playback function, an image playback function, etc.), etc.; the storage data area may Store data (such as audio data, phone book, etc.) created based on the use of mobile phones. In addition, the memory 12 may include a high-speed random access memory, and may also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), and a Secure Digital (SD) Card, Flash Card, at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage device.

The memory 12 may be an external memory and/or an internal memory of the electronic device 1. Further, the memory 12 may be a circuit with a storage function without a physical form in an integrated circuit, such as RAM (Random-Access Memory, random access memory), FIFO (First In First Out), etc. Alternatively, the memory 12 may also be a memory in physical form, such as a memory stick, a TF card (Trans-flash Card), and so on.

If the integrated module/unit of the electronic device 1 is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, this application implements all or part of the processes in the above-mentioned embodiments and methods, and can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the program is executed by the processor, the steps of the foregoing method embodiments can be implemented.

Wherein, the computer program includes computer program code, and the computer program code may be in the form of source code, object code, executable file, or some intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory) , Random Access Memory (RAM, Random Access Memory), electrical carrier signal, telecommunications signal and software distribution media, etc. It should be noted that the content contained in the computer-readable medium can be appropriately added or deleted in accordance with the requirements of the legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to the legislation and patent practice, the computer-readable medium Does not include electrical carrier signals and telecommunication signals.

With reference to Figure 1, the memory 12 in the electronic device 1 stores multiple instructions to implement a behavior prediction method based on a predictive model, and the processor 13 can execute the multiple instructions to implement: When the instruction is recognized, obtain at least one video to be detected; extract the target video sequence of each video to be detected in the at least one video to be detected; input each target video sequence into a pre-trained prediction model to determine the target person Fraud probability in each target video sequence; Based on the quantile principle, combine each fraud probability into a target vector; Input the target vector into a pre-trained classifier to determine the target probability; According to the target Probability, to determine whether the target person is at risk of fraud.

The embodiments of the present application also provide a computer-readable storage medium, the computer-readable storage medium stores a computer program, the computer program includes program instructions, and the processor executes the program instructions to implement the present application Any of the behavior prediction methods based on the prediction model provided in the embodiments.

The computer-readable storage medium may be the internal storage unit of the computer device described in the foregoing embodiment, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (SMC), or a secure digital (Secure Digital, SD) equipped on the computer device. ) Card, Flash Card, etc.

In the several embodiments provided in this application, it should be understood that the disclosed system, device, and method may be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

The modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.

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

For those skilled in the art, it is obvious that the present application is not limited to the details of the foregoing exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or basic characteristics of the application.

Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of this application is defined by the appended claims rather than the above description, and therefore it is intended to fall into the claims. All changes in the meaning and scope of the equivalent elements of are included in this application. Any associated diagram marks in the claims should not be regarded as limiting the claims involved.

In addition, it is obvious that the word "including" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claims can also be implemented by one unit or device through software or hardware. The second class words are used to indicate names, and do not indicate any specific order.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the application and not to limit them. Although the application has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the application can be Modifications or equivalent replacements are made without departing from the spirit and scope of the technical solution of this application.

Claims

A behavior prediction method based on a prediction model, the method comprising:

When receiving a fraud identification instruction, obtain at least one video to be detected;

Extracting the target video sequence of each video to be detected in the at least one video to be detected;

Input each target video sequence into a pre-trained prediction model to determine the fraud probability of the target person in each target video sequence;

Based on the quantile principle, combine each fraud probability into a target vector;

Input the target vector into a pre-trained classifier to determine the target probability;

According to the target probability, it is determined whether the target person has a risk of fraud.
The behavior prediction method based on a prediction model according to claim 1, wherein the at least one video to be detected includes one or a combination of the following:

During the bank business interview process, the answer video of each question; and/or

During the face-to-face review of the insurance business, the answer video for each question.
The behavior prediction method based on a prediction model according to claim 1, wherein said extracting said at least one video to be detected, the target video sequence of each video to be detected comprises:

A K-means clustering algorithm is used to extract the target video sequence of each of the at least one to-be-detected videos.
The behavior prediction method based on a prediction model according to claim 1, wherein, before each target video sequence is input into a pre-trained prediction model, the fraud probability of the target person in each target video sequence is determined. The method also includes:

Obtain sample video sequences;

Using a face recognition algorithm to extract the facial movements and eyeball angles of each person in the sample video sequence;

Use support vector regression algorithm to train the facial movements of each character to obtain the first vector of each character;

Training the eyeball angles of all the characters using a neural network algorithm to obtain the second vector of each character;

Intercept the head rotation angles of all the characters to obtain the third vector of each character;

Combine the first vector, second vector, and third vector of each character to obtain a sample vector of each character;

From the configuration database, determine the overdue data of each character to formulate the overdue label of each character;

The sample vector and the overdue label of each person are used as sample data, and the prediction model is trained using a gradient boosting algorithm.
The behavior prediction method based on a prediction model according to claim 1, wherein the determining whether the target person is at risk of fraud according to the target probability comprises:

When the target probability is greater than or equal to a preset threshold, it is determined that the target person is at risk of fraud; or

When the target probability is less than the preset threshold, it is determined that the target person has no risk of fraud.
The behavior prediction method based on the prediction model according to claim 1, wherein, when it is determined that the target person is at risk of fraud, the method further comprises:

Save the video corresponding to the target person;

Intercept the image of the target person from the saved video;

Combine the captured images and send prompt information to the designated terminal device.
The behavior prediction method based on the prediction model of claim 6, wherein when it is determined that the target person is at risk of fraud, the method further comprises:

Obtain all record information of the target person from the configuration database;

Sending all the recorded information to the designated terminal device.
A behavior prediction device based on a prediction model, the device comprising:

The obtaining unit is configured to obtain at least one video to be detected when a fraud identification instruction is received;

An extraction unit, configured to extract the target video sequence of each video to be detected in the at least one video to be detected;

The determining unit is used to input each target video sequence into a pre-trained prediction model to determine the fraud probability of the target person in each target video sequence;

The combination unit is used to combine each fraud probability into a target vector based on the quantile principle;

The determining unit is further configured to input the target vector into a pre-trained classifier to determine the target probability;

The determining unit is further configured to determine whether the target person has a risk of fraud according to the target probability.
An electronic device including a memory and a processor;

The memory is used to store computer readable instructions;

The processor is configured to execute the computer program and implement the following steps when executing the computer program:

When receiving a fraud identification instruction, obtain at least one video to be detected;

Extracting the target video sequence of each video to be detected in the at least one video to be detected;

Input each target video sequence into a pre-trained prediction model to determine the fraud probability of the target person in each target video sequence;

Based on the quantile principle, combine each fraud probability into a target vector;

Input the target vector into a pre-trained classifier to determine the target probability;

According to the target probability, it is determined whether the target person has a risk of fraud.
The electronic device according to claim 9, wherein the at least one video to be detected comprises one or a combination of the following:

During the bank business interview process, the answer video of each question; and/or

During the face-to-face review of the insurance business, the answer video for each question.
The electronic device according to claim 9, wherein said extracting the target video sequence of each of the at least one video to be detected in the at least one video to be detected by the processor comprises:

A K-means clustering algorithm is used to extract the target video sequence of each of the at least one to-be-detected videos.
The electronic device according to claim 9, wherein the processor implements the step of inputting each target video sequence into a pre-trained prediction model to determine the fraud probability of the target person in each target video sequence. ,Also includes:

Obtain sample video sequences;

Using a face recognition algorithm to extract the facial movements and eyeball angles of each person in the sample video sequence;

Use support vector regression algorithm to train the facial movements of each character to obtain the first vector of each character;

Training the eyeball angles of all the characters using a neural network algorithm to obtain the second vector of each character;

Intercept the head rotation angles of all the characters to obtain the third vector of each character;

Combine the first vector, second vector, and third vector of each character to obtain a sample vector of each character;

From the configuration database, determine the overdue data for each character to formulate the overdue label for each character;

The sample vector and the overdue label of each person are used as sample data, and the prediction model is trained using a gradient boosting algorithm.
9. The electronic device according to claim 9, wherein the processor to determine whether the target person is at risk of fraud according to the target probability comprises:

When the target probability is greater than or equal to a preset threshold, it is determined that the target person is at risk of fraud; or

When the target probability is less than the preset threshold, it is determined that the target person has no risk of fraud.
The electronic device according to claim 9, wherein the processor implementing when determining that the target person is at risk of fraud comprises:

Save the video corresponding to the target person;

Intercept the image of the target person from the saved video;

Combine the captured images and send prompt information to the designated terminal device.
The electronic device according to claim 14, wherein the processor implementing when determining that the target person is at risk of fraud comprises:

Obtain all record information of the target person from the configuration database;

Sending all the recorded information to the designated terminal device.
A computer-readable storage medium, the computer-readable storage medium stores at least one instruction, and the at least one instruction is executed by a processor in an electronic device to implement the method according to any one of claims 1 to 7 Behavior prediction method based on prediction model.
15. The computer-readable storage medium according to claim 16, wherein said extracting the target video sequence of each of the at least one video to be detected in the at least one video to be detected by the processor comprises:

A K-means clustering algorithm is used to extract the target video sequence of each of the at least one to-be-detected videos.
The computer-readable storage medium according to claim 16, wherein the processor implements the input of each target video sequence into a pre-trained prediction model to determine the position of the target person in each target video sequence. Before the probability of fraud, it also includes:

Obtain sample video sequences;

Using a face recognition algorithm to extract the facial movements and eyeball angles of each person in the sample video sequence;

Use support vector regression algorithm to train the facial movements of each character to obtain the first vector of each character;

Training the eyeball angles of all the characters using a neural network algorithm to obtain the second vector of each character;

Intercept the head rotation angles of all the characters to obtain the third vector of each character;

Combine the first vector, second vector, and third vector of each character to obtain a sample vector of each character;

From the configuration database, determine the overdue data of each character to formulate the overdue label of each character;

The sample vector and the overdue label of each person are used as sample data, and the prediction model is trained using a gradient boosting algorithm.
16. The computer-readable storage medium according to claim 16, wherein the processor implementing the determining whether the target person is at risk of fraud according to the target probability comprises:

When the target probability is greater than or equal to a preset threshold, it is determined that the target person is at risk of fraud; or

When the target probability is less than the preset threshold, it is determined that the target person has no risk of fraud.
15. The computer-readable storage medium according to claim 16, wherein the processor implementing when determining that the target person is at risk of fraud comprises:

Save the video corresponding to the target person;

Intercept the image of the target person from the saved video;

Combine the captured images and send prompt information to the designated terminal device.