CN117764710A - Monitoring method for housing financial risk behaviors - Google Patents
Monitoring method for housing financial risk behaviors Download PDFInfo
- Publication number
- CN117764710A CN117764710A CN202311748999.3A CN202311748999A CN117764710A CN 117764710 A CN117764710 A CN 117764710A CN 202311748999 A CN202311748999 A CN 202311748999A CN 117764710 A CN117764710 A CN 117764710A
- Authority
- CN
- China
- Prior art keywords
- information
- question
- housing
- income
- interview
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000006399 behavior Effects 0.000 title claims abstract description 102
- 238000012544 monitoring process Methods 0.000 title claims abstract description 64
- 238000000034 method Methods 0.000 title claims abstract description 23
- 238000000605 extraction Methods 0.000 claims description 105
- 230000007774 longterm Effects 0.000 claims description 26
- 230000010354 integration Effects 0.000 claims description 5
- 238000012550 audit Methods 0.000 claims description 4
- 230000004927 fusion Effects 0.000 claims description 3
- 238000012795 verification Methods 0.000 description 20
- 230000009471 action Effects 0.000 description 16
- 230000006870 function Effects 0.000 description 6
- 230000008569 process Effects 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000012417 linear regression Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 230000008439 repair process Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
Landscapes
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The invention relates to the field of housing finance, and discloses a monitoring method of housing finance risk behaviors, which comprises the following steps: generating an authenticity grading value according to the income information and the income certificate; if the value is larger than a first preset scoring threshold value, generating a predicted income amount; if the difference value between the declaration income amount and the predicted income amount is larger than or equal to a preset difference value threshold value, generating first-level risk behavior prompt information and interview questioning contents; performing voice recognition on the interview video to obtain interview text and a time axis; text matching is carried out on interview text and interview question contents, and a question-answering group sequence is obtained; generating text features and image features corresponding to each question-answering group and fusing the text features and the image features to obtain question-answering features; and generating risk behavior monitoring information according to the question-answering characteristics and the manual scores. Therefore, the auditing efficiency is improved, the fund default risk is reduced, and the fund safety of a financial institution is ensured.
Description
Technical Field
The invention relates to the field of housing finance, in particular to a monitoring method of housing finance risk behaviors.
Background
In the house property transaction process, the house property purchaser can relieve the fund pressure through personal house loans, long-term house deposit and other financial products. With the continuous development of the loan business of the individual housing, the risks facing the business are increasingly prominent, and the risk of the financial risk of the housing needs to be continuously monitored in each link of loan issuing.
In the existing process of monitoring the house financial risk behaviors, the following technical problems often exist:
firstly, the income proving materials submitted by borrowers are verified manually, so that time and labor are consumed, and depending on manual experience, omission easily occurs, so that house financial risk behaviors (such as fraud behaviors) are difficult to discover in time in the auditing process, thereby increasing the fund default risk and affecting the fund safety of financial institutions;
second, the establishment of long-term housing deposit provides a powerful guarantee for solving the housing problems of residents. However, the behavior of illegal long-term housing deposit frequently occurs, which causes the fund output risk of long-term housing deposit, and the effective means for monitoring the behavior of illegal long-term housing deposit is still lacking at present;
third, when verifying the application for long-term housing deposit extraction of the user, there is often a problem that the user uses the history photograph as an auxiliary proof image, thereby illegally extracting the long-term housing deposit and further causing a fund output risk of the long-term housing deposit.
Disclosure of Invention
This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter.
The invention provides a monitoring method for housing financial risk behaviors, which aims to solve one or more of the technical problems mentioned in the background art section.
The invention provides a monitoring method of housing financial risk behaviors, which comprises the following steps: the borrower who receives the housing loan uploads the income information and the income certificate through the terminal, and generates an authenticity score value corresponding to the income certificate according to the file type of the income certificate and a plurality of quality indexes of the income certificate, wherein the income information comprises the declaration income amount; if the authenticity score value is larger than a first preset score threshold value, obtaining borrower characteristic information of the borrower through the terminal, and inputting the borrower characteristic information into a pre-trained income prediction model to obtain predicted income amount of the borrower; if the difference value between the declaration income amount and the predicted income amount is larger than or equal to a preset difference value threshold, generating first-level risk behavior prompt information and interview questioning content corresponding to the borrower, and sending the first-level risk behavior prompt information and the interview questioning content to an auditing terminal so that auditing personnel corresponding to the auditing terminal interview with the borrower according to the interview questioning content; receiving interview videos and interview scores uploaded by auditors through audit terminals, wherein the interview scores comprise manual scores for each interview in interview questioning contents; performing voice recognition on the interview video to obtain interview text and a time axis corresponding to the interview text; text matching is carried out on interview text and interview question content to obtain a question and answer group sequence, wherein each question and answer group in the question and answer group sequence comprises a question and an answer; word embedding is carried out on each question-answer group, and text features corresponding to each question-answer group are obtained; determining a target video frame corresponding to each question-answer group in the interview video according to the question-answer group sequence and the time axis, and extracting features of the target video frame to obtain image features corresponding to each question-answer group; for each question-answering group, carrying out feature fusion on the corresponding text features and the corresponding image features to obtain corresponding question-answering features; and generating risk behavior monitoring information corresponding to the borrower according to the question answering characteristics corresponding to each question answering group and the manual scores of each question.
Optionally, the revenue prediction model includes a time sequence prediction sub-model, a regression sub-model and an integration sub-model, the revenue information includes a first revenue value sequence, and the first revenue value sequence includes revenue values respectively corresponding to a plurality of time points in a first historical time interval; and inputting borrower characteristic information into a pre-trained income prediction model to obtain predicted income amount of the borrower, comprising: inputting borrower characteristic information into a regression sub-model to obtain a first prediction income value; acquiring a second income value sequence, inputting the second income value sequence into a time sequence predictor model, and obtaining a second predicted income value corresponding to a target time point, wherein the target time point is a time point in a first historical time interval; inputting the first predicted income value and the second predicted income value into an integrated sub-model to obtain predicted income amount; and if the difference between the declared income amount and the predicted income amount is greater than or equal to a preset difference threshold, generating first-level risk behavior prompt information and interview question contents corresponding to the borrower, wherein the first-level risk behavior prompt information and interview question contents comprise: and determining the average declaration income according to the first income value sequence and the declaration income, and generating first-level risk behavior prompt information and interview question contents corresponding to the borrower if the difference between the average declaration income and the predicted income is greater than or equal to a preset difference threshold.
Optionally, according to the question-answer characteristics corresponding to each question-answer group and the manual score of each question, risk behavior monitoring information corresponding to the borrower is generated, including: according to the manual scores of questions included in each question-answer group, generating a weight corresponding to each question-answer group; generating a distinguishing feature according to the weight and the question-answer feature corresponding to each question-answer group; and inputting the discrimination characteristics into a risk discrimination network to obtain risk behavior monitoring information corresponding to the borrower.
Optionally, the monitoring method of the housing financial risk behavior further includes: when an extraction application for long-term housing deposit sent by a user through terminal equipment is received, determining whether an extraction reason category included in the extraction application is a first category, if the extraction reason category is the first category, acquiring a historical extraction evidence submitted by the user at each historical time point in a plurality of historical time points, and obtaining a historical extraction evidence set; carrying out house information identification on each history extraction proof file in the history extraction proof file set to obtain a house information set, and determining target house information from the house information set according to the corresponding history time point; determining whether the housing information contained in the extraction application is consistent with the target housing information, and if the housing information is consistent with the target housing information and the number of the history extraction certificates in the history extraction certificate set is greater than a preset number, controlling the terminal equipment to acquire auxiliary certificate images; generating extraction behavior monitoring information corresponding to the user according to the auxiliary proof image, wherein the extraction behavior monitoring information characterizes whether the extraction behavior of the user aiming at the long-term housing deposit is at risk or not; if the extraction behavior monitoring information characterizes that the user risks the extraction behavior of the long-term housing deposit, generating and sending prompt information characterizing that the extraction application does not pass to the terminal equipment.
Optionally, if the housing information is consistent with the target housing information and the number of the history extraction documents in the history extraction document set is greater than a preset number, the control terminal device collects the auxiliary document image, including: if the housing information is consistent with the target housing information and the number of the history extraction certificates in the history extraction certificate set is larger than the preset number, the control terminal device executes the following image acquisition operation: collecting a house image as an auxiliary proof image; generating a housing image file according to the current position information, the current time information and the auxiliary proof image, and uploading the housing image file; and generating the extracted behavior monitoring information corresponding to the user according to the auxiliary proof image, comprising: and generating the extracted behavior monitoring information corresponding to the user according to the housing image file.
Optionally, if the housing information is consistent with the target housing information and the number of the history extraction documents in the history extraction document set is greater than a preset number, the control terminal device collects the auxiliary document image, including: if the housing information is consistent with the target housing information and the number of the history extraction certificates in the history extraction certificate set is larger than the preset number, the control terminal device executes the following image acquisition operation: collecting a house image as an auxiliary proof image; generating a random code; generating a housing image file according to the current position information, the current time information, the random code and the auxiliary proof image, and uploading the housing image file; and generating the extracted behavior monitoring information corresponding to the user according to the auxiliary proof image, comprising: and generating the extracted behavior monitoring information corresponding to the user according to the housing image file.
The invention has the following beneficial effects: borrower risk behavior identification is achieved through automatic scoring and revenue prediction of revenue certificates. And the income proving material submitted by the borrower is not required to be manually verified, so that the auditing efficiency is improved. When the predicted income amount and the declared income amount are greatly different, risk behavior monitoring information is further generated according to interview videos in combination with manual scoring, and combination of manual evaluation and intelligent recognition is realized, so that auditing accuracy is improved, missing of house financial risk behaviors (such as fraudulent behaviors) is avoided, fund default risks are reduced, and fund safety of financial institutions is guaranteed.
Drawings
The above and other features, advantages and aspects of embodiments of the present invention will become more apparent by reference to the following detailed description when taken in conjunction with the accompanying drawings. The same or similar reference numbers will be used throughout the drawings to refer to the same or like elements. It should be understood that the figures are schematic and that elements and components are not necessarily drawn to scale.
Fig. 1 is a flow chart of a method of monitoring housing financial risk behavior in accordance with the present invention.
Detailed Description
The invention will be described in more detail below with reference to the accompanying drawings. While certain embodiments of the invention have been illustrated in the accompanying drawings, it is to be understood that the invention may be embodied in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete. It should be understood that the drawings and embodiments of the invention are for illustration purposes only and are not intended to limit the scope of the present invention.
It should be noted that, for convenience of description, only the portions related to the present invention are shown in the drawings. Embodiments of the invention and features of the embodiments may be combined with each other without conflict.
It should be noted that the terms "first," "second," and the like herein are merely used for distinguishing between different devices, modules, or units and not for limiting the order or interdependence of the functions performed by such devices, modules, or units.
It should be noted that references to "one", "a plurality" and "a plurality" in this disclosure are intended to be illustrative rather than limiting, and those skilled in the art will appreciate that "one or more" is intended to be construed as "one or more" unless the context clearly indicates otherwise.
The names of messages or information interacted between the devices of the present invention are for illustrative purposes only and are not intended to limit the scope of such messages or information.
The invention will be described in detail below with reference to the drawings in connection with embodiments.
As shown in fig. 1, a flowchart of a method for monitoring a housing financial risk behavior according to the present invention is shown, comprising the steps of:
step 101, receiving income information and income certificate uploaded by borrowers of housing loans through terminals, and generating an authenticity score value corresponding to the income certificate according to the file type of the income certificate and a plurality of quality indexes of the income certificate, wherein the income information comprises a declaration income amount.
In some embodiments, an execution subject of a monitoring method of housing financial risk behavior of the present invention is a monitoring backend server of a financial institution. The execution subject can first receive the income information and income certificate uploaded by the terminal. The income information includes the declaration income amount, and can be the average income amount in a certain time interval (for example, half a year) according to requirements. Optionally, the revenue information may also include revenue types, which may include, but are not limited to: payroll, business revenue, labor revenue, and the like. Accordingly, the revenue certificate may be a payroll certificate, a personal account billing detail file, etc. issued by the unit, and may be divided into different categories according to the unit of issue, for example, the file categories may include, but are not limited to: third party institution certificates, bank certificates, work unit certificates, etc., different file categories are configured with different scoring weights. The revenue document may be an image, and the plurality of quality indicators of the revenue document include, but are not limited to: image definition, seal definition in an image, and the like. And then, determining corresponding sub-grading values according to the index value of each quality index in the plurality of quality indexes, adding the sub-grading values to obtain candidate grading values, and multiplying the candidate grading values by grading weights corresponding to the file types to obtain the authenticity grading values corresponding to the income-proving files.
Step 102, if the authenticity score value is greater than a first preset score threshold value, obtaining borrower characteristic information of the borrower through the terminal, and inputting the borrower characteristic information into a pre-trained income prediction model to obtain predicted income amount of the borrower.
In some embodiments, if the authenticity score is greater than the first preset score threshold, the executing body may obtain borrower feature information of the borrower through the terminal. The borrower characteristic information includes, but is not limited to, the following dimensions: age, marital status, academic, job title, employer information, working years, number of households, guaranty category, etc. On this basis, borrower characteristic information is input into a pre-trained revenue prediction model. The income prediction model can be a multiple linear regression model, takes the borrower characteristic information as input, and outputs predicted income amount.
Optionally, the revenue prediction model may further include a time series prediction sub-model, a regression sub-model, and an integration sub-model, wherein the time series prediction sub-model may be a long short term memory network (LSTM). The regression sub-model may be a multiple linear regression model, and the regression sub-model takes the borrower characteristic information as input to output a first income prediction value. The integrated submodel may be an elastic network (elastic net), which is a linear regression model, which is a combination of ridge regression and Lasso regression. On the basis, the borrower characteristic information can be input into a regression sub-model to obtain a first prediction income value; acquiring a second income value sequence, inputting the second income value sequence into a time sequence predictor model, and obtaining a second predicted income value corresponding to a target time point, wherein the target time point is a time point in a first historical time interval; inputting the first predicted income value and the second predicted income value into an integrated sub-model to obtain predicted income amount; and determining the average declaration income according to the first income value sequence and the declaration income, and generating first-level risk behavior prompt information and interview question contents corresponding to the borrower if the difference between the average declaration income and the predicted income is greater than or equal to a preset difference threshold. The first income value sequence comprises income values corresponding to a plurality of time points in the first historical time interval. The first historical time interval is a first time interval before the current time point, the second historical time interval is a second time interval before the current time point, and the second historical time interval is a time interval before the first historical time interval. For example, the current time point is 12 months, the first historical time interval is 6 months before to 12 months, at this time, the first revenue value sequence is a sequence consisting of revenue values of each month from 6 months before to 12 months, the second historical time interval is 1 month to 5 months, and the second revenue value sequence is a sequence consisting of revenue values of 1 month to 5 months. On the basis, the target time point is any time point in the first historical time interval. Further, the prediction income amount is obtained by inputting the first prediction income value and the second prediction income value into the integration sub-model, wherein the prediction accuracy can be improved due to the integration of the prediction results of the regression sub-model and the time sequence prediction sub-model.
On the basis, the average declaration income amount is compared with the predicted income amount, and if the difference value is larger than or equal to a preset difference value threshold (for example, 2000), first-level risk behavior prompt information and interview question contents corresponding to the borrower are generated.
And 103, if the difference value between the declaration income amount and the predicted income amount is greater than or equal to a preset difference value threshold, generating first-level risk behavior prompt information and interview questioning contents corresponding to the borrower, and sending the first-level risk behavior prompt information and the interview questioning contents to the auditing terminal so that the auditing personnel corresponding to the auditing terminal interviews with the borrower according to the interview questioning contents.
In some embodiments, if the difference between the declared and predicted revenue amounts is greater than or equal to a preset difference threshold, the executing entity may generate first-level risk behavior prompt information and interview question contents corresponding to the borrower. The first-level risk behavior prompt information characterizes false risks of reporting income amount. On the basis, interview question contents are generated according to the borrower characteristic information. Therefore, the auditor can conduct interview with the borrower according to interview questioning content, and in the process, the interview process is shot through the shooting equipment, so that interview videos are obtained. In the process, after the auditor can ask questions, each question can be scored according to answers of borrowers, so that the manual score of each question is obtained and uploaded to the execution body through the audit terminal.
Wherein, the executing body can generate interview question contents in the following way: and acquiring borrower characteristic information, and taking each dimension included in the borrower characteristic information as input of a interview question content generation network to obtain interview question content. The interview question content generation network may be a text generation network (e.g., RNN, generation countermeasure network, etc.), and training is performed by a sample set to obtain the interview question content generation network. The samples in the sample set comprise sample borrower characteristic information and sample interview questioning contents. Then training is performed by means of machine learning (back propagation, random gradient descent) to obtain the interview question content generation network.
And 104, receiving interview videos and interview scores uploaded by auditors through the auditing terminal, wherein the interview scores comprise manual scores for each question in interview question contents.
In some embodiments, the executing entity may receive interview videos and interview scores uploaded by the auditor through the audit terminal.
Step 105, performing voice recognition on the interview video to obtain interview text and a time axis corresponding to the interview text; text matching is carried out on interview text and interview question content to obtain a question and answer group sequence, wherein each question and answer group in the question and answer group sequence comprises a question and an answer; word embedding is carried out on each question-answer group, and text features corresponding to each question-answer group are obtained; determining a target video frame corresponding to each question-answer group in the interview video according to the question-answer group sequence and the time axis, and extracting features of the target video frame to obtain image features corresponding to each question-answer group; and for each question-answering group, carrying out feature fusion on the corresponding text features and the corresponding image features to obtain the corresponding question-answering features.
In some embodiments, the executing entity may perform voice recognition on the interview video through a voice recognition technology to obtain interview text and a time axis corresponding to the interview text. And then, text matching can be carried out on interview texts and interview question contents, so that answers corresponding to each question are obtained, one question and one answer form a question-answer group, and a plurality of question-answer groups are arranged according to time sequence, so that a question-answer group sequence is obtained. On the basis, for each question-answer group, a target time point corresponding to a time axis is determined, at least one video frame corresponding to the target time point is extracted from interview video, and key frames are selected from the at least one video frame as target video frames. And inputting the target video frame into a convolutional neural network to obtain the image characteristics corresponding to each question-answer group. And for each question-answering group, splicing the corresponding text feature and the corresponding image feature to obtain the corresponding question-answering feature.
And 106, generating risk behavior monitoring information corresponding to the borrower according to the question-answer characteristics corresponding to each question-answer group and the manual scores of each question.
In some embodiments, the question and answer features and the manual score of each question are input into a risk discrimination network (such as a recurrent neural network, a decision tree, etc.), and risk behavior monitoring information corresponding to the borrower is generated. The risk behavior monitoring information is used for representing whether the borrowing behavior of the borrower is a risk behavior or not. On the basis of adopting a model structure of the cyclic neural network, training can be performed by using a training sample set, so that a risk discrimination network is obtained. The training samples comprise sample question-answer features, sample manual scores and corresponding sample risk behavior monitoring information. And the difference between the actual output and the expected output is determined by taking the sample question-answer characteristics and the sample manual scores as inputs and the corresponding sample risk behavior monitoring information as the expected output, and then the difference is reversely transmitted into the circulating neural network by using a reverse propagation algorithm, and the parameters of the circulating neural network are adjusted until the training stopping condition is met, so that the risk judging network is obtained.
Optionally, according to the question-answer characteristics corresponding to each question-answer group and the manual score of each question, risk behavior monitoring information corresponding to the borrower is generated, including: according to the manual scores of questions included in each question-answer group, generating a weight corresponding to each question-answer group; generating a distinguishing feature according to the weight and the question-answer feature corresponding to each question-answer group; and inputting the discrimination characteristics into a risk discrimination network to obtain risk behavior monitoring information corresponding to the borrower. The weight corresponding to each question-answer group can be obtained through a preset weight conversion formula, and generally, the higher the manual score is, the larger the corresponding weight is. And then taking the first non-zero value in the weight corresponding to each question-answer group as the target times, copying the question-answer characteristics corresponding to the question-answer groups for the target times to obtain the copied characteristics, splicing the copied characteristics corresponding to each question-answer group to obtain the distinguishing characteristics, and finally inputting the distinguishing characteristics into a risk distinguishing network to obtain the risk behavior monitoring information corresponding to the borrower.
In some embodiments, borrower risk performance identification is achieved by automatic scoring and revenue prediction of revenue certificates. The income proof material submitted by borrowers does not need to be manually verified, so that the auditing efficiency is improved. When the predicted income amount and the declared income amount are greatly different, risk behavior monitoring information is further generated according to interview videos in combination with manual scoring, and combination of manual evaluation and intelligent recognition is realized, so that auditing accuracy is improved, missing of house financial risk behaviors (such as fraudulent behaviors) is avoided, fund default risks are reduced, and fund safety of financial institutions is guaranteed.
In some embodiments, in order to further solve the second technical problem described in the background section, namely "lack of effective means for monitoring behaviors of illegal long-term housing deposit", a method for monitoring behaviors of housing financial risk further includes the following steps:
step one, when an extraction application for long-term housing cash is received, which is sent by a user through a terminal device, determining whether an extraction reason category included in the extraction application is a first category, and if the extraction reason category is the first category (for example, a house renting), acquiring a history extraction proof file submitted by the user at each history time point in a plurality of history time points, and obtaining a history extraction proof file set. In practice, according to different uses of extracting long-term housing deposit, the extraction reasons can be divided into different categories, and the extraction reason categories can include: purchase, lease, repair, illness, etc. Different extraction cause categories correspond to different extraction documents. For example, when the first category is a rental room, the extraction proof file and the extraction proof file are both rental room contracts.
And thirdly, carrying out house information identification on each history extraction proof file in the history extraction proof file set to obtain a house information set, and determining target house information from the house information set according to the corresponding history time point. The housing information may include housing position information, house type information, and the like. The historical extraction proof files can be identified by utilizing an OCR technology, and then the preset keywords are used for searching, so that housing information corresponding to each historical extraction proof file is obtained, and a housing information set is obtained. The history extraction document corresponds to a history time point, and correspondingly, housing information extracted from the history extraction document corresponds to a history time point. The target housing information may be housing information having a latest historical time point corresponding to the housing information set.
And step four, determining whether the housing information included in the extraction application is consistent with the target housing information, and if the housing information is consistent with the target housing information and the number of the history extraction evidence files in the history extraction evidence file set is greater than the preset number, controlling the terminal equipment to acquire the auxiliary evidence images. Wherein the preset number (e.g., 3) may be determined based on the local average room change period. When the housing information is consistent with the target housing information, the user can be considered to rent the same house in the year and when the user submits the extraction application last time, and the user submits the extraction application for a plurality of times (the number of times of the extraction application is larger than the preset number), so that the risk degree of the extraction action aiming at the long-term housing deposit is higher. Thus, further verification by the auxiliary proof image is required.
And fifthly, generating extraction behavior monitoring information corresponding to the user according to the auxiliary proof image, wherein the extraction behavior monitoring information characterizes whether the extraction behavior of the user aiming at the long-term housing deposit is at risk or not. Specifically, whether the auxiliary proof image meets a preset verification condition set can be determined, and if so, extraction behavior monitoring information which characterizes that the extraction behavior is risk-free is generated; conversely, extraction behavior monitoring information that characterizes the risk of extraction behavior may be generated.
Step six, if the extraction behavior monitoring information characterizes that the user risks the extraction behavior of the long-term housing deposit, generating and sending prompt information characterizing that the extraction application does not pass to the terminal equipment.
Optionally, if the housing information is consistent with the target housing information and the number of the history extraction documents in the history extraction document set is greater than a preset number, the control terminal device collects the auxiliary document image, including: if the housing information is consistent with the target housing information and the number of the history extraction certificates in the history extraction certificate set is larger than the preset number, the control terminal device executes the following image acquisition operation: collecting a house image as an auxiliary proof image; generating a housing image file according to the current position information, the current time information and the auxiliary proof image, and uploading the housing image file; and generating the extracted behavior monitoring information corresponding to the user according to the auxiliary proof image, comprising: and generating the extracted behavior monitoring information corresponding to the user according to the housing image file. The current position information and the current time information can be used as attribute information of the auxiliary proof image, and then a housing image file is formed.
Optionally, if the housing information is consistent with the target housing information and the number of the history extraction documents in the history extraction document set is greater than a preset number, the control terminal device collects the auxiliary document image, including: if the housing information is consistent with the target housing information and the number of the history extraction certificates in the history extraction certificate set is larger than the preset number, the control terminal device executes the following image acquisition operation: collecting a house image as an auxiliary proof image; generating a random code; generating a housing image file according to the current position information, the current time information, the random code and the auxiliary proof image, and uploading the housing image file; and generating the extracted behavior monitoring information corresponding to the user according to the auxiliary proof image, comprising: and generating the extracted behavior monitoring information corresponding to the user according to the housing image file. The current position information, the current time information and the random code can be used as attribute information of the auxiliary proof image, so that a housing image file is formed. Meanwhile, the execution subject can generate the random code based on the current time by adopting the same hash function, and the execution subject and the terminal equipment adopt the same hash function and generate the random code based on the same time, so that the obtained random codes are the same. On the basis, the execution main body can analyze the house image file on the basis of receiving the house image file, obtain the random code therein and compare the random code with the random code locally stored by the execution main body so as to verify the house image file, prevent a user from falsifying the house image file and ensure the authenticity of the auxiliary proof image.
In the embodiments, the extraction application with higher risk degree is further verified through the auxiliary proof image, so that high-risk extraction behaviors are avoided, further, behaviors of illegal extraction of long-term housing deposit are effectively monitored, and the fund output risk of long-term housing deposit is reduced.
In some embodiments, in order to further solve the third technical problem described in the background section, that is, "when verifying an application for extracting long-term housing deposit from a user, there is often a problem that the user uses a history photo as an auxiliary proof image, thereby illicitly extracting the long-term housing deposit and further creating a fund output risk of the long-term housing deposit", in some embodiments of the present invention, if housing information is consistent with target housing information and the number of history extraction proof files in a history extraction proof file set is greater than a preset number, the control terminal device collects the auxiliary proof image, including the following steps:
step one, if the housing information is consistent with the target housing information and the number of the history extraction certificates in the history extraction certificate set is larger than the preset number, determining whether the housing information comprises a housing pattern, and if not, sending a housing pattern uploading prompt message to the terminal equipment;
step two, obtaining a house type diagram through terminal equipment, marking a plurality of verification positions in the house type diagram, determining corresponding verification action information from a pre-configured verification action information base for each verification position, and sending the marked house type diagram and the verification action information corresponding to each verification position to the terminal equipment; the terminal device performs the following at each verification location: collecting images of local houses and verification actions corresponding to the verification positions as auxiliary proof images; generating a random code; and generating a housing image file according to the current position information, the current time information, the random code and the auxiliary proof image, and uploading the housing image file. Thus, the housing image files corresponding to the verification positions respectively form a housing image file set. It will be appreciated that the secondary proof image includes both a house (e.g., kitchen) at the verification location and a verification action (e.g., standing) corresponding to the verification location.
Further, according to the auxiliary proof image, generating the extracted behavior monitoring information corresponding to the user includes: and generating the extracted behavior monitoring information corresponding to the user according to the housing image file set. Specifically, the executing body can analyze the house image file to obtain the current position information, the current time information, the random code and the auxiliary proof image, and inquire the corresponding three-dimensional model according to the house information. On the basis, the execution main body can analyze the house image file on the basis of receiving the house image file, obtain the random code therein and compare the random code with the random code locally stored by the execution main body so as to verify the house image file, prevent a user from falsifying the house image file and ensure the authenticity of the auxiliary proof image. In practice, the executing body may use the same hash function to generate the random code based on the current time, and since the executing body and the terminal device use the same hash function and generate the random code based on the same time, the obtained random codes are the same. In addition, the auxiliary proof image of each verification location may be verified. Specifically, the auxiliary proof image can be input into the action recognition network to obtain a corresponding recognition action. Then, according to the verification position (for example, kitchen), locating a local space corresponding to the auxiliary proof image from the three-dimensional model, and comparing the image of the local space with the auxiliary proof image to determine whether the image is the same space; meanwhile, comparing the identification action with the corresponding verification action, and if the comparison is consistent and the same space is determined, generating extraction behavior monitoring information which characterizes that the extraction behavior is not at risk; conversely, extraction behavior monitoring information that characterizes the risk of extraction behavior may be generated. The motion recognition network may be various image recognition networks, such as convolutional neural networks, and is trained using a set of motion samples. The action samples in the action sample set include action sample images and corresponding recognition actions (annotated action names). The similarity between the two images can be calculated by comparing the image of the local space with the auxiliary proof image, and the two images can be considered as the same space when the similarity is larger than a preset threshold (for example, 90%).
In these embodiments, by combining the verification location with the verification action, the user is prevented from utilizing the existing historical photographs as auxiliary proof images, the security of long-term housing fund storage is improved, and the fund output risk is reduced.
The above description is only illustrative of the few preferred embodiments of the present invention and of the principles of the technology employed. It will be appreciated by persons skilled in the art that the scope of the invention referred to in the present invention is not limited to the specific combinations of the technical features described above, but also covers other technical features formed by any combination of the technical features described above or their equivalents without departing from the inventive concept described above. Such as the above-mentioned features and the technical features disclosed in the present invention (but not limited to) having similar functions are replaced with each other.
Claims (6)
1. A method for monitoring financial risk behavior of a housing, comprising:
receiving income information and income certificate uploaded by a borrower of housing loan through a terminal, and generating an authenticity credit value corresponding to the income certificate according to the file type of the income certificate and a plurality of quality indexes of the income certificate, wherein the income information comprises a declaration income amount;
if the authenticity score value is larger than a first preset score threshold value, obtaining borrower characteristic information of the borrower through the terminal, and inputting the borrower characteristic information into a pre-trained income prediction model to obtain predicted income amount of the borrower;
if the difference value between the declaration income amount and the predicted income amount is larger than or equal to a preset difference value threshold, generating first-level risk behavior prompt information and interview questioning content corresponding to the borrower, and sending the first-level risk behavior prompt information and the interview questioning content to an auditing terminal so that an auditing person corresponding to the auditing terminal interviews with the borrower according to the interview questioning content;
receiving interview videos and interview scores uploaded by the auditor through the audit terminal, wherein the interview scores comprise manual scores for each question in the interview question content;
performing voice recognition on the interview video to obtain interview texts and time axes corresponding to the interview texts; text matching is carried out on the interview text and the interview question content to obtain a question and answer group sequence, wherein each question and answer group in the question and answer group sequence comprises a question and an answer; word embedding is carried out on each question-answer group, and text features corresponding to each question-answer group are obtained; determining a target video frame corresponding to each question-answer group in the interview video according to the question-answer group sequence and the time axis, and extracting features of the target video frame to obtain image features corresponding to each question-answer group; for each question-answering group, carrying out feature fusion on the corresponding text features and the corresponding image features to obtain corresponding question-answering features;
and generating risk behavior monitoring information corresponding to the borrower according to the question-answer characteristics corresponding to each question-answer group and the manual scores of each question.
2. The method of claim 1, wherein the revenue prediction model comprises a time series prediction sub-model, a regression sub-model, and an integration sub-model, the revenue information comprises a first revenue value sequence comprising revenue values corresponding to a plurality of time points within a first historical time interval, respectively; and
the step of inputting the borrower characteristic information into a pre-trained income prediction model to obtain the predicted income amount of the borrower, comprising the following steps:
inputting the borrower characteristic information into the regression sub-model to obtain a first prediction income value;
acquiring a second income value sequence, and inputting the second income value sequence into the time sequence prediction sub-model to obtain a second predicted income value corresponding to a target time point, wherein the target time point is a time point in the first historical time interval;
inputting the first predicted income value and the second predicted income value into the integrated submodel to obtain the predicted income amount; and
and if the difference value between the declaration income amount and the predicted income amount is greater than or equal to a preset difference value threshold, generating first-level risk behavior prompt information and interview questioning contents corresponding to the borrower, wherein the first-level risk behavior prompt information and interview questioning contents comprise:
and determining a declaration average income according to the first income value sequence and the declaration income, and generating first-level risk behavior prompt information and interview questioning contents corresponding to the borrower if the difference value between the declaration average income and the predicted income is greater than or equal to a preset difference value threshold.
3. The method for monitoring the financial risk behaviors of the housing according to claim 2, wherein the generating risk behavior monitoring information corresponding to the borrower according to the question-answer characteristics corresponding to each question-answer group and the manual score of each question comprises:
according to the manual scores of questions included in each question-answer group, generating weights corresponding to each question-answer group;
generating a distinguishing feature according to the weight and the question-answer feature corresponding to each question-answer group;
and inputting the distinguishing characteristics into a risk distinguishing network to obtain risk behavior monitoring information corresponding to the borrower.
4. The method for monitoring residential financial risk behaviors according to claim 3, further comprising:
when an extraction application for long-term housing deposit sent by a user through terminal equipment is received, determining whether an extraction reason category included in the extraction application is a first category, if the extraction reason category is the first category, acquiring a history extraction evidence submitted by the user at each history time point in a plurality of history time points, and obtaining a history extraction evidence set;
carrying out house information identification on each history extraction proof file in the history extraction proof file set to obtain a house information set, and determining target house information from the house information set according to a corresponding history time point;
determining whether housing information included in the extraction application is consistent with the target housing information, and if the housing information is consistent with the target housing information and the number of the history extraction evidence files in the history extraction evidence file set is greater than a preset number, controlling the terminal equipment to acquire an auxiliary evidence image;
generating extraction behavior monitoring information corresponding to the user according to the auxiliary proof image, wherein the extraction behavior monitoring information characterizes whether the extraction behavior of the user aiming at long-term housing deposit is at risk or not;
if the extraction behavior monitoring information characterizes that the user risks the extraction behavior of the long-term housing deposit, generating and sending prompt information characterizing that the extraction application does not pass to the terminal equipment.
5. The method for monitoring a housing financial risk behavior according to claim 4, wherein if the housing information is consistent with the target housing information and the number of the history extraction documents in the history extraction document set is greater than a preset number, controlling the terminal device to collect an auxiliary document image, comprises:
if the housing information is consistent with the target housing information and the number of the history extraction certificates in the history extraction certificate set is larger than the preset number, controlling the terminal equipment to execute the following image acquisition operation: collecting a house image as the auxiliary proof image; generating a housing image file according to the current position information, the current time information and the auxiliary proof image, and uploading the housing image file; and
the generating the extracted behavior monitoring information corresponding to the user according to the auxiliary proof image comprises the following steps:
and generating the extracted behavior monitoring information corresponding to the user according to the housing image file.
6. The method for monitoring a housing financial risk behavior according to claim 4, wherein if the housing information is consistent with the target housing information and the number of the history extraction documents in the history extraction document set is greater than a preset number, controlling the terminal device to collect an auxiliary document image, comprises:
if the housing information is consistent with the target housing information and the number of the history extraction certificates in the history extraction certificate set is larger than the preset number, controlling the terminal equipment to execute the following image acquisition operation: collecting a house image as the auxiliary proof image; generating a random code; generating a housing image file according to the current position information, the current time information, the random code and the auxiliary proof image, and uploading the housing image file; and
the generating the extracted behavior monitoring information corresponding to the user according to the auxiliary proof image comprises the following steps:
and generating the extracted behavior monitoring information corresponding to the user according to the housing image file.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311748999.3A CN117764710B (en) | 2023-12-19 | 2023-12-19 | Monitoring method for housing financial risk behaviors |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311748999.3A CN117764710B (en) | 2023-12-19 | 2023-12-19 | Monitoring method for housing financial risk behaviors |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117764710A true CN117764710A (en) | 2024-03-26 |
| CN117764710B CN117764710B (en) | 2024-06-04 |
Family
ID=90311663
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311748999.3A Active CN117764710B (en) | 2023-12-19 | 2023-12-19 | Monitoring method for housing financial risk behaviors |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117764710B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6178945B1 (en) * | 2017-05-26 | 2017-08-09 | 株式会社野村総合研究所 | Fractional funds transfer accumulation system, program and method |
| CN109034742A (en) * | 2018-07-19 | 2018-12-18 | 武汉杰威鑫创软件技术有限公司 | A kind of public accumalation fund for housing construction archive management system |
| CN109697665A (en) * | 2018-12-15 | 2019-04-30 | 深圳壹账通智能科技有限公司 | Loan checking method, device, equipment and medium based on artificial intelligence |
| CN112102069A (en) * | 2020-09-18 | 2020-12-18 | 华院分析技术(上海)有限公司 | Personal property mortgage loan information input analysis system |
| US20210173854A1 (en) * | 2019-12-05 | 2021-06-10 | Murray B. WILSHINSKY | Method and system for self-aggregation of personal data and control thereof |
| CN113298373A (en) * | 2021-05-20 | 2021-08-24 | 中国建设银行股份有限公司 | Financial risk assessment method, device, storage medium and equipment |
| CN114219604A (en) * | 2021-11-09 | 2022-03-22 | 中国建设银行股份有限公司 | House loan service processing method, device, equipment and storage medium |
| CN117035986A (en) * | 2023-08-01 | 2023-11-10 | 中国建设银行股份有限公司 | Remote double-record loan method and device, electronic equipment and storage medium |
-
2023
- 2023-12-19 CN CN202311748999.3A patent/CN117764710B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP6178945B1 (en) * | 2017-05-26 | 2017-08-09 | 株式会社野村総合研究所 | Fractional funds transfer accumulation system, program and method |
| CN109034742A (en) * | 2018-07-19 | 2018-12-18 | 武汉杰威鑫创软件技术有限公司 | A kind of public accumalation fund for housing construction archive management system |
| CN109697665A (en) * | 2018-12-15 | 2019-04-30 | 深圳壹账通智能科技有限公司 | Loan checking method, device, equipment and medium based on artificial intelligence |
| US20210173854A1 (en) * | 2019-12-05 | 2021-06-10 | Murray B. WILSHINSKY | Method and system for self-aggregation of personal data and control thereof |
| CN112102069A (en) * | 2020-09-18 | 2020-12-18 | 华院分析技术(上海)有限公司 | Personal property mortgage loan information input analysis system |
| CN113298373A (en) * | 2021-05-20 | 2021-08-24 | 中国建设银行股份有限公司 | Financial risk assessment method, device, storage medium and equipment |
| CN114219604A (en) * | 2021-11-09 | 2022-03-22 | 中国建设银行股份有限公司 | House loan service processing method, device, equipment and storage medium |
| CN117035986A (en) * | 2023-08-01 | 2023-11-10 | 中国建设银行股份有限公司 | Remote double-record loan method and device, electronic equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 王慧: "公积金贷款及提取的风险防范", 中国市场, no. 27, 30 September 2021 (2021-09-30), pages 43 - 44 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117764710B (en) | 2024-06-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Cai et al. | Judging online peer-to-peer lending behavior: A comparison of first-time and repeated borrowing requests | |
| CN109255499B (en) | Complaint and complaint case processing method, device and equipment | |
| US20180260891A1 (en) | Systems and methods for generating and using optimized ensemble models | |
| CN113011973B (en) | Method and equipment for financial transaction supervision model based on intelligent contract data lake | |
| CN112102076B (en) | Comprehensive risk early warning system of platform | |
| CN113014566B (en) | Malicious registration detection method and device, computer readable medium and electronic device | |
| CN108961032A (en) | Borrow or lend money processing method, device and server | |
| Liu et al. | Can listing information indicate borrower credit risk in online peer-to-peer lending? | |
| Xu et al. | Identifying features for detecting fraudulent loan requests on P2P platforms | |
| CN113888368A (en) | Feature selection method for criminal case detention risk assessment based on image coding | |
| CN113159796A (en) | Trade contract verification method and device | |
| Xu et al. | PEER-TO-PEER LOAN FRAUD DETECTION: CONSTRUCTING FEATURES FROM TRANSACTION DATA. | |
| CN117764710B (en) | Monitoring method for housing financial risk behaviors | |
| Yangyudongnanxin | Financial credit risk control strategy based on weighted random forest algorithm | |
| CN116342141A (en) | Method, device and equipment for identifying empty shell enterprises | |
| CN115907968A (en) | Wind control rejection inference method and device based on pedestrian credit | |
| CN113706300A (en) | Loan method and device for small and micro enterprises | |
| CN115564591A (en) | Financing product determination method and related equipment | |
| CN110766091B (en) | Method and system for identifying trepanning loan group partner | |
| Hackney et al. | Automating Sanctions Compliance: Aligning Regulatory Technology, Rules and Goals | |
| Langenbucher | Consumer credit in the age of AI–beyond anti-discrimination law | |
| Locatelli et al. | Artificial Intelligence and Credit Risk | |
| CN116070641B (en) | Online interpretation method of electronic contract | |
| CN117114844A (en) | Information verification method, apparatus, computer device and storage medium | |
| CN113283977A (en) | Credit scoring decision analysis method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |