CN117710096B - Post-loan risk monitoring method and system for agricultural loan based on land parcel information - Google Patents
Post-loan risk monitoring method and system for agricultural loan based on land parcel information Download PDFInfo
- Publication number
- CN117710096B CN117710096B CN202410161725.2A CN202410161725A CN117710096B CN 117710096 B CN117710096 B CN 117710096B CN 202410161725 A CN202410161725 A CN 202410161725A CN 117710096 B CN117710096 B CN 117710096B
- Authority
- CN
- China
- Prior art keywords
- loan
- semantic
- loan application
- feature map
- self
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 75
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000012545 processing Methods 0.000 claims abstract description 10
- 230000003993 interaction Effects 0.000 claims description 102
- 238000012549 training Methods 0.000 claims description 57
- 239000013598 vector Substances 0.000 claims description 52
- 230000003044 adaptive effect Effects 0.000 claims description 19
- 238000012482 interaction analysis Methods 0.000 claims description 14
- 238000013527 convolutional neural network Methods 0.000 claims description 11
- 238000003062 neural network model Methods 0.000 claims description 11
- 238000005728 strengthening Methods 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 7
- 238000012935 Averaging Methods 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 3
- 238000011835 investigation Methods 0.000 abstract description 11
- 238000001514 detection method Methods 0.000 abstract description 4
- 238000009826 distribution Methods 0.000 description 18
- 238000012271 agricultural production Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 239000002689 soil Substances 0.000 description 3
- 238000007792 addition Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000002787 reinforcement Effects 0.000 description 2
- 238000012502 risk assessment Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000003337 fertilizer Substances 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 239000000575 pesticide Substances 0.000 description 1
- 238000005215 recombination Methods 0.000 description 1
- 230000006798 recombination Effects 0.000 description 1
Landscapes
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
A post-loan risk monitoring method and system of agricultural loans based on land parcel information relates to the technical field of intelligent loan risk monitoring, and the post-loan risk monitoring method and system are characterized in that remote sensing images of loan application information of target customers and remote sensing images of target land parcel are acquired, and then image processing and semantic understanding algorithms are introduced into the rear end to analyze the remote sensing images and the loan application information, so that post-loan risk detection and grade judgment are comprehensively carried out, the accuracy and efficiency of post-loan risk monitoring can be improved, the cost and risk loss of manual investigation are reduced, and better decision basis is provided for financial institutions.
Description
Technical Field
The application relates to the technical field of intelligent loan risk monitoring, in particular to a post-loan risk monitoring method and system for an agricultural loan based on land parcel information.
Background
Agricultural loans refer to loans that a financial institution provides to an agricultural production operator for agricultural production, operations, or rural construction. Agricultural loans are characterized by a shorter loan period, a smaller loan amount, a dispersed loan object, and a higher risk after lending. To effectively control the risk of agricultural loans, financial institutions need post-loan risk monitoring of the credit status of the loan clients and the value of the mortgage.
However, the conventional post-loan risk monitoring method is mainly based on the repayment condition and the financial index of the borrower, but the method often cannot fully understand the actual operation condition and risk condition of the borrower. In addition, some post-loan risk monitoring methods also typically rely on manual investigation in the field, which is time consuming, labor intensive, costly, inefficient, and difficult to cover for all loan customers and mortgages. Therefore, how to utilize modern information technology to improve the monitoring efficiency and accuracy after loan of agricultural loans and reduce the monitoring cost and risk loss is a problem to be solved urgently.
Accordingly, a post-loan risk monitoring scheme for agricultural loans based on parcel information is desired.
Disclosure of Invention
The present application has been made to solve the above-mentioned technical problems. The application provides a post-loan risk monitoring method and system for agricultural loans based on land information, which are characterized in that by collecting loan application information of a target customer and remote sensing images of a target land, wherein the loan application information comprises project information of a credit project of the target customer and land information of a mortgage target land in the credit project, and then, an image processing and semantic understanding algorithm is introduced at the rear end to analyze the remote sensing images and the loan application information, so that post-loan risk detection and grade judgment are comprehensively carried out, the accuracy and efficiency of post-loan risk monitoring can be improved, the cost and risk loss of manual investigation are reduced, and a better decision basis is provided for financial institutions.
In a first aspect, there is provided a post-loan risk monitoring method of an agricultural loan based on parcel information, comprising:
Obtaining loan application information of a target client;
acquiring a remote sensing image of a target land block;
Carrying out semantic coding on the loan application information to obtain a loan application information semantic coding feature vector;
extracting features of the remote sensing image by a land parcel element semantic feature extractor based on a deep neural network model to obtain a remote sensing image land parcel element semantic feature map;
Performing cross-modal semantic interaction analysis on the remote sensing image parcel element semantic feature map and the loan application information semantic coding feature vector to obtain a loan application-parcel element semantic interaction feature map;
The loan application-land parcel element semantic interaction feature map is used for obtaining an adaptive enhanced loan application-land parcel element semantic interaction feature map through an adaptive attention module to serve as adaptive enhanced loan application-land parcel element semantic interaction feature;
And determining a post-loan risk level label based on the self-adaptive enhanced loan application-land parcel element semantic interaction characteristics.
In a second aspect, there is provided a post-loan risk monitoring system for an agricultural loan based on plot information, comprising:
The loan application information acquisition module is used for acquiring loan application information of the target client;
the remote sensing image acquisition module is used for acquiring a remote sensing image of the target land block;
The semantic coding module is used for carrying out semantic coding on the loan application information to obtain a loan application information semantic coding feature vector;
the feature extraction module is used for extracting features of the remote sensing image through a land parcel element semantic feature extractor based on a deep neural network model so as to obtain a remote sensing image land parcel element semantic feature map;
The cross-modal semantic interaction analysis module is used for carrying out cross-modal semantic interaction analysis on the remote sensing image land parcel element semantic feature map and the loan application information semantic coding feature vector so as to obtain a loan application-land parcel element semantic interaction feature map;
The self-adaptive attention module is used for enabling the loan application-land parcel element semantic interaction feature map to pass through the self-adaptive attention module to obtain a self-adaptive enhanced loan application-land parcel element semantic interaction feature map as self-adaptive enhanced loan application-land parcel element semantic interaction features;
And the post-loan risk level label determining module is used for determining a post-loan risk level label based on the self-adaptive enhanced loan application-land parcel element semantic interaction characteristics.
Compared with the prior art, the post-loan risk monitoring method and system for the agricultural loan based on the land information are characterized in that the post-loan risk is detected and judged comprehensively by collecting the loan application information of the target client and the remote sensing image of the target land, wherein the loan application information comprises project information of a credit project of the target client and land information of the target land which is mortgage in the credit project, and then the analysis of the remote sensing image and the loan application information is carried out by introducing an image processing and semantic understanding algorithm at the rear end, so that the accuracy and the efficiency of post-loan risk monitoring can be improved, the cost and the risk loss of manual investigation are reduced, and a better decision basis is provided for a financial institution.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a post-loan risk monitoring method of an agricultural loan based on parcel information, in accordance with an embodiment of the application.
Fig. 2 is a schematic diagram of a post-loan risk monitoring method of an agricultural loan based on plot information, according to an embodiment of the application.
Fig. 3 is a block diagram of a post-loan risk monitoring system for an agricultural loan based on parcel information, in accordance with an embodiment of the application.
Fig. 4 is a schematic view of a scenario of a post-loan risk monitoring method of an agricultural loan based on land parcel information, according to an embodiment of the application.
Detailed Description
The following description of the technical solutions according to the embodiments of the present application will be given with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
Unless defined otherwise, all technical and scientific terms used in the embodiments of the application have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the present application is for the purpose of describing particular embodiments only and is not intended to limit the scope of the present application.
In describing embodiments of the present application, unless otherwise indicated and limited thereto, the term "connected" should be construed broadly, for example, it may be an electrical connection, or may be a communication between two elements, or may be a direct connection, or may be an indirect connection via an intermediate medium, and it will be understood by those skilled in the art that the specific meaning of the term may be interpreted according to circumstances.
It should be noted that, the term "first\second\third" related to the embodiment of the present application is merely to distinguish similar objects, and does not represent a specific order for the objects, it is to be understood that "first\second\third" may interchange a specific order or sequence where allowed. It is to be understood that the "first\second\third" distinguishing objects may be interchanged where appropriate such that embodiments of the application described herein may be practiced in sequences other than those illustrated or described herein.
Agricultural loans refer to loans that a financial institution provides to an agricultural production operator for agricultural production, operations, or rural construction, such loans typically being used to purchase seeds, fertilizers, pesticides, agricultural machinery, pay for labor costs, and other expenses related to agricultural production.
Agricultural production cycles are relatively short, so agricultural loans typically have a short loan term to accommodate the characteristics of agricultural production; agricultural loan amounts are typically small relative to loans for some large industrial or commercial projects, as the capital requirements of agricultural production operators are typically relatively limited; borrowers of agricultural loans are typically decentralized agricultural production operators, rather than being concentrated on a small number of large enterprises or institutions; because agricultural production is greatly influenced by factors such as weather, natural disasters and the like, and factors such as price fluctuation of agricultural products and the like, the risk of agricultural loans after lending is relatively high.
To effectively control the risk of agricultural loans, financial institutions often need to take a number of measures, including but not limited to:
Assessment of credit status financial institutions assess credit status prior to issuing loans to agricultural production operators to ensure borrowers have the ability to repay the loans.
Mortgage assessment, when issuing agricultural loans, financial institutions often require borrowers to offer mortgages as guarantees of the loan, and thus the value of the mortgage needs to be assessed.
Post-loan risk monitoring, once a loan is issued, financial institutions need to monitor the loan clients' business status, repayment status, and mortgage value periodically, and discover and deal with possible risks in time.
Traditional post-loan risk monitoring methods rely primarily on loan customers' repayment and financial indicators, and generally include: monitoring whether a loan client pays on time or not and whether overdue pays exist or not; evaluating the operating status and repayment capacity of the loan client through financial statements and financial indicators (such as a liability statement, profit statement, cash flow statement, etc.); sometimes, a field investigation needs to be performed manually to understand the actual operation situation of the loan clients, the situation of the mortgage, etc.
However, the conventional post-loan risk monitoring method has some drawbacks, and the conventional method mainly depends on financial reports and repayment records provided by customers, so that actual business situations and risk situations of the customers cannot be comprehensively known, and the customers may provide incomplete or inaccurate information, so that monitoring results are not accurate enough. Monitoring methods that rely on payment records and financial statements are generally periodic and do not know in real time the business condition and risk changes of the customer, so the monitoring results may be delayed. In the traditional post-loan monitoring method, the field investigation needs to consume a large amount of manpower and material resources, has high cost, and is difficult to cover all loan clients and mortgages. The traditional method is difficult to discover risk signals in advance, once problems are serious, measures cannot be taken in time to prevent risks. Relying on manual field investigation is susceptible to subjective factors and there may be errors in the monitoring results.
Therefore, how to utilize modern information technology to improve the monitoring efficiency and accuracy after loan of agricultural loans and reduce the monitoring cost and risk loss is a current urgent problem to be solved.
Fig. 1 is a flowchart of a post-loan risk monitoring method of an agricultural loan based on parcel information, in accordance with an embodiment of the application. Fig. 2 is a schematic diagram of a post-loan risk monitoring method of an agricultural loan based on plot information, according to an embodiment of the application. As shown in fig. 1 and 2, the post-loan risk monitoring method of the agricultural loan based on the plot information includes: 110, obtaining loan application information of a target client; 120, acquiring a remote sensing image of the target land block; 130, performing semantic coding on the loan application information to obtain a loan application information semantic coding feature vector; 140, extracting features of the remote sensing image through a land parcel element semantic feature extractor based on a deep neural network model to obtain a remote sensing image land parcel element semantic feature map; 150, performing cross-modal semantic interaction analysis on the remote sensing image parcel element semantic feature map and the loan application information semantic coding feature vector to obtain a loan application-parcel element semantic interaction feature map; 160, obtaining an adaptive enhanced loan application-parcel element semantic interaction feature map as an adaptive enhanced loan application-parcel element semantic interaction feature by the adaptive attention module from the loan application-parcel element semantic interaction feature map; and 170, determining a post-loan risk level tag based on the self-adaptive enhanced loan application-land parcel element semantic interaction characteristics.
The deep neural network model is a convolutional neural network model.
With the development of remote sensing technology and artificial intelligence, post-loan risk monitoring of agricultural loans is possible by utilizing land parcel information and remote sensing images. The land parcel information comprises the position, the area, the soil type, the crop planting condition and the like of the land parcel, and the remote sensing image can provide real-time monitoring and evaluation data of the land parcel. By combining the land parcel information and the remote sensing image, the management risk of the borrower and the risk condition of the loan can be detected more accurately. Based on the above, in the technical scheme of the application, a post-loan risk monitoring method of an agricultural loan based on land parcel information is provided, which can collect loan application information of a target customer and remote sensing images of the target land parcel, wherein the loan application information comprises project information of a credit project of the target customer and land parcel information of a mortgage target land parcel in the credit project, and then image processing and semantic understanding algorithm are introduced at the rear end to analyze the remote sensing images and the loan application information, so as to comprehensively detect post-loan risk and judge the grade, thereby improving accuracy and efficiency of post-loan risk monitoring, reducing cost and risk loss of manual investigation, and providing better decision basis for financial institutions.
Specifically, in the technical scheme of the application, first, loan application information of a target client and a remote sensing image of a target land block are acquired. It should be appreciated that the loan application information includes project information for a target customer's credit project and parcel information for a mortgage target parcel in the credit project. Both of these pieces of information are important components of post-loan risk monitoring. The project information of the target customer's credit project may provide important information about the borrower's basic condition, the loan amount, the loan period, etc. Such information may reflect the credit status and repayment capabilities of the borrower, playing an important role in assessing the risk of the loan. The land block information of the mortgage target land block in the credit project provides specific data about the position, the area, the soil type, the crop planting condition and the like of the land block. Such parcel information may help assess the borrower's agricultural business, land value, and mortgage value, as well as make sense for risk assessment of the loan. Therefore, in order to perform semantic analysis and understanding on the loan application information so as to convert the key semantics into feature vectors to assist in subsequent data analysis and risk monitoring, in the technical scheme of the application, the loan application information needs to be subjected to semantic coding to obtain the loan application information semantic coding feature vectors. By semantically encoding the loan application information, the situation of the borrower and the value of the mortgage can be more comprehensively known, so that the accuracy and the efficiency of post-loan risk monitoring are improved.
Then, in order to extract the semantic features of the land parcel elements from the remote sensing image for subsequent data processing and analysis, feature mining is carried out on the remote sensing image by a land parcel element semantic feature extractor based on a convolutional neural network model, wherein the land parcel element semantic feature extractor has excellent performance in the aspect of implicit feature extraction of the image, so that land parcel element semantic feature information about a target land parcel in the remote sensing image is extracted, and a remote sensing image land parcel element semantic feature map is obtained. In particular, the land element semantic feature extractor may learn feature information about the texture, shape, color, etc. of the target land in the remote sensing image, and semantic feature information about land elements, such as vegetation coverage, soil texture, crop type, etc. These features may help to better understand the characteristics and status of the plot and provide more information for subsequent post-loan risk assessment.
It should be appreciated that the remote sensing image parcel element semantic feature map includes element semantic features related to the target parcel, and the loan application information semantic coding feature vector includes text information features of the loan application. In the technical scheme of the application, the remote sensing image land parcel element semantic feature map and the loan application information semantic coding feature vector are further processed through a cross-modal semantic interactor based on a meta-network model to obtain a loan application-land parcel element semantic interaction feature map. By processing the cross-modal semantic interactors based on the meta-network model, information interaction and fusion among different modal data can be realized, so that more comprehensive and more accurate post-credit risk characteristic representation is provided. This helps to improve the accuracy and effectiveness of post-loan risk monitoring, thereby providing a more reliable basis for decision making for financial institutions.
In a specific embodiment of the present application, performing cross-modal semantic interaction analysis on the remote sensing image parcel element semantic feature map and the loan application information semantic coding feature vector to obtain a loan application-parcel element semantic interaction feature map, including: and the remote sensing image land parcel element semantic feature map and the loan application information semantic coding feature vector are passed through a cross-mode semantic interactor based on a meta-network model to obtain the loan application-land parcel element semantic interaction feature map.
Further, the remote sensing image land parcel element semantic feature map and the loan application information semantic coding feature vector are passed through a cross-modal semantic interactor based on a meta-network model to obtain the loan application-land parcel element semantic interaction feature map, which comprises the following steps: the remote sensing image land parcel element semantic feature map is passed through a convolutional neural network of the cross-modal semantic interactor based on a meta-network model to obtain a remote sensing image land parcel feature map; the loan application information semantic coding feature vector passes through a one-dimensional convolution layer of the cross-modal semantic interactor based on the meta-network model to obtain a loan application information association feature vector; and carrying out channel weighting on the remote sensing image parcel feature map by using the loan application information associated feature vector so as to obtain the loan application-parcel element semantic interaction feature map.
Further, it is contemplated that each channel of the loan application-parcel element semantic interaction feature map represents a different semantic interaction feature, however, not all features are of equal importance for post-loan risk monitoring. That is, some features may be more critical, better reflecting the risk of loans, and some may be extraneous interfering features. Therefore, in order to be able to further strengthen the key semantic interaction features between loan application-parcel elements, it is necessary to focus attention on important channel features, reducing the impact of extraneous features on post-loan risk monitoring. Based on the above, in the technical scheme of the application, important information in the feature map can be weighted through an attention mechanism so as to improve the perception and recognition capability of key features. Specifically, in the technical scheme of the application, the loan application-parcel element semantic interaction feature map is further processed through an adaptive attention module to obtain an adaptive enhanced loan application-parcel element semantic interaction feature map. It should be appreciated that the adaptive attention module converts the feature map of each channel into a weight value by using a meta-weight generator. The weight values are multiplied by the loan application-land parcel element semantic interaction feature map channel by channel, so that each channel in the feature map receives different degrees of attention, and important channel feature information related to risk monitoring after loan is highlighted. In this way, attention can be focused on important features, and the influence of irrelevant features is weakened, so that the post-credit risk level judging capability and the classifying capability of the classifier are improved.
In a specific embodiment of the present application, the loan application-parcel element semantic interaction feature map is used as an adaptive enhanced loan application-parcel element semantic interaction feature by an adaptive attention module, and the method includes: the loan application-parcel element semantic interaction feature map is subjected to the following self-adaptive strengthening formula through a self-adaptive attention module to obtain the self-adaptive strengthening loan application-parcel element semantic interaction feature map; wherein, the self-adaptive strengthening formula is:
Wherein is the loan application-parcel element semantic interaction feature map, wherein/> represents global averaging processing of each feature matrix along a channel dimension in the feature map, wherein/> is a channel feature vector of the loan application-parcel element semantic interaction feature map,/> and/> are weights and offsets of a convolution layer,/> is an activation function,/> is a convolution feature vector of the channel feature vector,/> is a feature value of each position in the convolution feature vector,/> is a weight vector,/> is a per-position point multiplier, and/> is the self-adaptive enhanced loan application-parcel element semantic interaction feature map.
And then, the self-adaptive enhanced loan application-plot element semantic interaction feature map is passed through a classifier to obtain a classification result, wherein the classification result is used for representing a post-loan risk level label. Specifically, the classification label of the classifier is a post-credit risk grade label, such as low risk, medium risk or high risk, so that the detection and grade judgment of post-credit risk can be comprehensively performed by using the classification result. Therefore, the accuracy and the efficiency of risk monitoring after lending can be improved, the cost and the risk loss of manual investigation are reduced, and a better decision basis is provided for financial institutions.
In one embodiment of the present application, determining a post-loan risk level tag based on the adaptively enhanced loan application-parcel element semantic interaction characteristics, comprises: and the self-adaptive enhanced loan application-land parcel element semantic interaction feature map is passed through a classifier to obtain a classification result, wherein the classification result is used for representing a post-loan risk level label.
In one embodiment of the present application, the post-loan risk monitoring method of the agricultural loan based on land parcel information further comprises a training step of: the device is used for training the land parcel element semantic feature extractor based on the convolutional neural network model, the cross-modal semantic interactor based on the meta-network model, the self-adaptive attention module and the classifier. The training step comprises the following steps: acquiring training loan application information of a target client; acquiring a training remote sensing image of a target land block; carrying out semantic coding on the training loan application information to obtain training loan application information semantic coding feature vectors; performing feature extraction on the training remote sensing image by using the land parcel element semantic feature extractor based on the deep neural network model to obtain a training remote sensing image land parcel element semantic feature map; performing cross-modal semantic interaction analysis on the training remote sensing image land parcel element semantic feature map and the training loan application information semantic coding feature vector to obtain a training loan application-land parcel element semantic interaction feature map; optimizing the training loan application-parcel element semantic interaction feature map to obtain an optimized training loan application-parcel element semantic interaction feature map; the optimized training loan application-land parcel element semantic interaction feature map is passed through the self-adaptive attention module to obtain a training self-adaptive enhanced loan application-land parcel element semantic interaction feature map; passing the training self-adaptive enhanced loan application-land parcel element semantic interaction feature map through the classifier to obtain a classification loss function value; training the land parcel element semantic feature extractor based on the convolutional neural network model, the cross-modal semantic interactor based on the meta-network model, the self-adaptive attention module and the classifier based on the classification loss function value.
In the above technical solution, each feature matrix of the training remote sensing image parcel element semantic feature map expresses an image semantic feature of the training remote sensing image of the target parcel, and each feature matrix follows a channel dimension distribution of the convolutional neural network model, so that after a cross-modal semantic interactor based on a meta-network model is used to perform cross-modal interaction on the training remote sensing image parcel element semantic feature map and the training loan application information semantic coding feature vector, the training remote sensing image parcel element semantic feature map is constrained in the channel dimension based on the coding semantic feature of the training loan application information of the target customer expressed by the training loan application information semantic coding feature vector, which also makes the training loan application-parcel element semantic interaction feature map have a mixed distribution weakened in association with the image semantic distribution of the feature matrix in the channel dimension, so that a probability density representation of the training application-parcel element semantic interaction feature map based on the image semantic feature space distribution and the channel dimension distribution is sparse in a probability density domain. And the training loan application-plot element semantic interaction feature map passes through the self-adaptive attention module, and then the self-adaptive attention module is used for carrying out self-adaptive reinforcement on each feature matrix of the training loan application-plot element semantic interaction feature map along the channel dimension, so that the probability density representation of the whole feature distribution of the training self-adaptive reinforcement loan application-plot element semantic interaction feature map based on the image semantic feature spatial distribution and the channel dimension distribution under the probability density domain is further thinned, and the classification result regression convergence effect is affected when the classification judgment is carried out through a classifier.
Based on this, the applicant of the present application optimizes the training loan application-parcel element semantic interaction feature map , expressed as: optimizing the training loan application-land parcel element semantic interaction feature map by using the following optimization formula to obtain an optimized training loan application-land parcel element semantic interaction feature map; wherein, the optimization formula is:
Wherein represents a position-by-position square graph of a feature graph,/> is a parameter trainable intermediate weight graph, for example, based on the image semantic feature spatial distribution and channel dimension distribution properties of the loan application-parcel element semantic interaction feature graph/> , the feature value of each feature matrix is initially set to be the feature value mean value of the corresponding feature matrix of the remote sensing image parcel element semantic feature graph, then the feature vectors are weighted along the channel by the loan application information semantic coding feature vector, in addition,/> is a single bitmap with all feature values of 1,/> is the training loan application-parcel element semantic interaction feature graph,/> is the optimized training loan application-parcel element semantic interaction feature graph,/> represents position-by-position addition and/> represents position-by-point multiplication.
Here, in order to optimize the distribution uniformity and consistency of the sparse probability density of the training loan application-plot element semantic interaction feature map in the overall probability space, the distance distribution optimization based on the spatial angle inclination is performed on the distance type spatial distribution of the training loan application-plot element semantic interaction feature map/> in the high-dimensional feature space through a tail distribution strengthening mechanism of a quasi-standard cauchy distribution type, so that the feature distribution spatial resonance of weak correlation of the distance of each local feature distribution of the training loan application-plot element semantic interaction feature map/> is realized, the uniformity and consistency of the training loan application-plot element semantic interaction feature map/> overall in the probability density distribution layer with respect to regression probability convergence is improved, and the classification regression convergence effect, namely the classification convergence speed and accuracy are improved. Therefore, the post-loan risk detection and grade judgment can be comprehensively carried out based on the land block information and the remote sensing image, so that the accuracy and the efficiency of post-loan risk monitoring are improved, the cost and the risk loss of manual investigation are reduced, and a better decision basis is provided for financial institutions.
In summary, the post-loan risk monitoring method of the agricultural loan based on the plot information is clarified, which can improve the accuracy and efficiency of post-loan risk monitoring, reduce the cost and risk loss of manual investigation and provide better decision basis for financial institutions.
In one embodiment of the application, FIG. 3 is a block diagram of a post-loan risk monitoring system for an agricultural loan based on parcel information, in accordance with an embodiment of the application. As shown in fig. 3, a post-loan risk monitoring system 200 of an agricultural loan based on land parcel information, according to an embodiment of the application, comprises: a loan application information obtaining module 210, configured to obtain loan application information of a target client; the remote sensing image acquisition module 220 is configured to acquire a remote sensing image of a target land block; the semantic coding module 230 is configured to perform semantic coding on the loan application information to obtain a loan application information semantic coding feature vector; the feature extraction module 240 is configured to perform feature extraction on the remote sensing image by using a land parcel feature semantic feature extractor based on a deep neural network model to obtain a remote sensing image land parcel feature semantic feature map; the cross-modal semantic interaction analysis module 250 is configured to perform cross-modal semantic interaction analysis on the remote sensing image parcel element semantic feature map and the loan application information semantic coding feature vector to obtain a loan application-parcel element semantic interaction feature map; the adaptive attention module 260 is configured to make the loan application-parcel element semantic interaction feature map pass through the adaptive attention module to obtain an adaptive enhanced loan application-parcel element semantic interaction feature map as an adaptive enhanced loan application-parcel element semantic interaction feature; the post-loan risk level tag determining module 270 is configured to determine a post-loan risk level tag based on the self-adaptive enhanced loan application-parcel element semantic interaction feature.
Here, it will be understood by those skilled in the art that the specific functions and operations of the respective units and modules in the above-described post-loan risk monitoring system for an agricultural loan based on plot information have been described in detail in the above description of the post-loan risk monitoring method for an agricultural loan based on plot information with reference to fig. 1 to 2, and thus, repeated descriptions thereof will be omitted.
As described above, the post-loan risk monitoring system 200 of an agricultural loan based on land parcel information according to an embodiment of the application may be implemented in various terminal devices, such as a server for post-loan risk monitoring of an agricultural loan based on land parcel information, and the like. In one example, the post-loan risk monitoring system 200 for an agricultural loan based on parcel information, in accordance with an embodiment of the application, may be integrated into the terminal device as a software module and/or hardware module. For example, the post-loan risk monitoring system 200 for land parcel information-based agricultural loans may be a software module in the operating system of the terminal device, or may be an application developed for the terminal device; of course, the post-loan risk monitoring system 200 for agricultural loans based on land parcel information may also be one of the plurality of hardware modules of the terminal device.
Alternatively, in another example, the post-loan risk monitoring system 200 of the agricultural loan based on the plot information and the terminal device may be separate devices, and the post-loan risk monitoring system 200 of the agricultural loan based on the plot information may be connected to the terminal device through a wired and/or wireless network and transmit the interaction information in a agreed data format.
Fig. 4 is a schematic view of a scenario of a post-loan risk monitoring method of an agricultural loan based on land parcel information, according to an embodiment of the application. As shown in fig. 4, in this application scenario, first, loan application information of a target client is acquired (e.g., C1 as illustrated in fig. 4); and, acquiring a remote sensing image of the target plot (e.g., C2 as illustrated in fig. 4); the acquired loan application information and remote sensing image are then input into a server (e.g., S as illustrated in fig. 4) that deploys a post-loan risk monitoring algorithm for the agricultural loan based on the plot information, wherein the server is capable of processing the loan application information and the remote sensing image by the post-loan risk monitoring algorithm for the agricultural loan based on the plot information to determine a post-loan risk level tag.
It is also noted that in the apparatus, devices and methods of the present application, the components or steps may be disassembled and/or assembled. Such decomposition and/or recombination should be considered as equivalent aspects of the present application.
The previous description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects without departing from the scope of the application. Thus, the present application is not intended to be limited to the aspects shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
The foregoing description has been presented for purposes of illustration and description. Furthermore, this description is not intended to limit embodiments of the application to the form disclosed herein. Although a number of example aspects and embodiments have been discussed above, a person of ordinary skill in the art will recognize certain variations, modifications, alterations, additions, and subcombinations thereof.
Claims (7)
1. A post-loan risk monitoring method for an agricultural loan based on parcel information, comprising:
Obtaining loan application information of a target client;
acquiring a remote sensing image of a target land block;
Carrying out semantic coding on the loan application information to obtain a loan application information semantic coding feature vector;
extracting features of the remote sensing image by a land parcel element semantic feature extractor based on a deep neural network model to obtain a remote sensing image land parcel element semantic feature map;
Performing cross-modal semantic interaction analysis on the remote sensing image parcel element semantic feature map and the loan application information semantic coding feature vector to obtain a loan application-parcel element semantic interaction feature map;
The loan application-land parcel element semantic interaction feature map is used for obtaining an adaptive enhanced loan application-land parcel element semantic interaction feature map through an adaptive attention module to serve as adaptive enhanced loan application-land parcel element semantic interaction feature;
Determining a post-loan risk level tag based on the self-adaptive enhanced loan application-parcel element semantic interaction features;
The cross-modal semantic interaction analysis is performed on the remote sensing image parcel element semantic feature map and the loan application information semantic coding feature vector to obtain a loan application-parcel element semantic interaction feature map, which comprises the following steps: the remote sensing image land parcel element semantic feature map and the loan application information semantic coding feature vector are passed through a cross-modal semantic interactor based on a meta-network model to obtain the loan application-land parcel element semantic interaction feature map;
The method for obtaining the self-adaptive enhanced loan application-parcel element semantic interaction feature map through the self-adaptive attention module, wherein the self-adaptive enhanced loan application-parcel element semantic interaction feature map is used as the self-adaptive enhanced loan application-parcel element semantic interaction feature, and comprises the following steps: the loan application-parcel element semantic interaction feature map is subjected to the following self-adaptive strengthening formula through a self-adaptive attention module to obtain the self-adaptive strengthening loan application-parcel element semantic interaction feature map;
wherein, the self-adaptive strengthening formula is:
；
；
；
；
Wherein is the loan application-parcel element semantic interaction feature map, wherein/> represents global averaging processing of each feature matrix along a channel dimension in the feature map, wherein/> is a channel feature vector of the loan application-parcel element semantic interaction feature map,/> and/> are weights and biases of a convolution layer,/> is an activation function,/> is a convolution feature vector of the channel feature vector,/> is a feature value of each position in the convolution feature vector,/> is a weight vector,/> is a per-position point multiplication, and/> is the self-adaptive enhanced loan application-parcel element semantic interaction feature map;
wherein determining a post-loan risk level tag based on the self-adaptive enhanced loan application-parcel feature semantic interaction feature comprises: and the self-adaptive enhanced loan application-land parcel element semantic interaction feature map is passed through a classifier to obtain a classification result, wherein the classification result is used for representing a post-loan risk level label.
2. The post-loan risk monitoring method of an agricultural loan based on plot information of claim 1, wherein the deep neural network model is a convolutional neural network model.
3. The post-loan risk monitoring method of an agricultural loan based on parcel information of claim 2, wherein passing the remote sensing image parcel element semantic feature map and the loan application information semantic coding feature vector through a meta-network model-based cross-modal semantic interactor to obtain the loan application-parcel element semantic interaction feature map, comprises:
The remote sensing image land parcel element semantic feature map is passed through a convolutional neural network of the cross-modal semantic interactor based on a meta-network model to obtain a remote sensing image land parcel feature map;
The loan application information semantic coding feature vector passes through a one-dimensional convolution layer of the cross-modal semantic interactor based on the meta-network model to obtain a loan application information association feature vector;
And carrying out channel weighting on the remote sensing image parcel feature map by using the loan application information associated feature vector so as to obtain the loan application-parcel element semantic interaction feature map.
4. The post-loan risk monitoring method of an agricultural loan based on land parcel information of claim 3, further comprising the training step of: the device is used for training a land parcel element semantic feature extractor based on a convolutional neural network model, the cross-modal semantic interactor based on a meta-network model, the self-adaptive attention module and the classifier.
5. The post-loan risk monitoring method of an agricultural loan based on land parcel information as recited in claim 4, wherein said training step comprises:
Acquiring training loan application information of a target client;
acquiring a training remote sensing image of a target land block;
Carrying out semantic coding on the training loan application information to obtain training loan application information semantic coding feature vectors;
Performing feature extraction on the training remote sensing image by using the land parcel element semantic feature extractor based on the deep neural network model to obtain a training remote sensing image land parcel element semantic feature map;
Performing cross-modal semantic interaction analysis on the training remote sensing image land parcel element semantic feature map and the training loan application information semantic coding feature vector to obtain a training loan application-land parcel element semantic interaction feature map;
Optimizing the training loan application-parcel element semantic interaction feature map to obtain an optimized training loan application-parcel element semantic interaction feature map;
the optimized training loan application-land parcel element semantic interaction feature map is passed through the self-adaptive attention module to obtain a training self-adaptive enhanced loan application-land parcel element semantic interaction feature map;
Passing the training self-adaptive enhanced loan application-land parcel element semantic interaction feature map through the classifier to obtain a classification loss function value;
Training the land parcel element semantic feature extractor based on the convolutional neural network model, the cross-modal semantic interactor based on the meta-network model, the self-adaptive attention module and the classifier based on the classification loss function value.
6. A post-loan risk monitoring system for an agricultural loan based on parcel information, comprising:
The loan application information acquisition module is used for acquiring loan application information of the target client;
the remote sensing image acquisition module is used for acquiring a remote sensing image of the target land block;
The semantic coding module is used for carrying out semantic coding on the loan application information to obtain a loan application information semantic coding feature vector;
the feature extraction module is used for extracting features of the remote sensing image through a land parcel element semantic feature extractor based on a deep neural network model so as to obtain a remote sensing image land parcel element semantic feature map;
The cross-modal semantic interaction analysis module is used for carrying out cross-modal semantic interaction analysis on the remote sensing image land parcel element semantic feature map and the loan application information semantic coding feature vector so as to obtain a loan application-land parcel element semantic interaction feature map;
The self-adaptive attention module is used for enabling the loan application-land parcel element semantic interaction feature map to pass through the self-adaptive attention module to obtain a self-adaptive enhanced loan application-land parcel element semantic interaction feature map as self-adaptive enhanced loan application-land parcel element semantic interaction features;
The post-loan risk level label determining module is used for determining a post-loan risk level label based on the self-adaptive enhanced loan application-land parcel element semantic interaction characteristics;
Wherein, cross-modal semantic interaction analysis module includes: the remote sensing image land parcel element semantic feature map and the loan application information semantic coding feature vector are passed through a cross-modal semantic interactor based on a meta-network model to obtain the loan application-land parcel element semantic interaction feature map;
Wherein the adaptive attention module comprises: the loan application-parcel element semantic interaction feature map is subjected to the following self-adaptive strengthening formula through a self-adaptive attention module to obtain the self-adaptive strengthening loan application-parcel element semantic interaction feature map;
wherein, the self-adaptive strengthening formula is:
；
；
；
；
wherein is the loan application-parcel element semantic interaction feature map, wherein/> represents global averaging processing of each feature matrix along a channel dimension in the feature map, wherein/> is a channel feature vector of the loan application-parcel element semantic interaction feature map,/> and/> are weights and biases of a convolution layer,/> is an activation function,/> is a convolution feature vector of the channel feature vector,/> is a feature value of each position in the convolution feature vector,/> is a weight vector,/> is a per-position point multiplication, and/> is the self-adaptive enhanced loan application-parcel element semantic interaction feature map;
The post-loan risk level label determining module comprises: and the self-adaptive enhanced loan application-land parcel element semantic interaction feature map is passed through a classifier to obtain a classification result, wherein the classification result is used for representing a post-loan risk level label.
7. The post-loan risk monitoring system of an agricultural loan based on plot information of claim 6, wherein the deep neural network model is a convolutional neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410161725.2A CN117710096B (en) | 2024-02-05 | 2024-02-05 | Post-loan risk monitoring method and system for agricultural loan based on land parcel information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410161725.2A CN117710096B (en) | 2024-02-05 | 2024-02-05 | Post-loan risk monitoring method and system for agricultural loan based on land parcel information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117710096A CN117710096A (en) | 2024-03-15 |
| CN117710096B true CN117710096B (en) | 2024-04-16 |
Family
ID=90162800
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410161725.2A Active CN117710096B (en) | 2024-02-05 | 2024-02-05 | Post-loan risk monitoring method and system for agricultural loan based on land parcel information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117710096B (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116310822A (en) * | 2023-03-09 | 2023-06-23 | 中国工商银行股份有限公司 | Loan request processing method and device for cultivated land and electronic equipment |
| CN116308754A (en) * | 2023-03-22 | 2023-06-23 | 广州信瑞泰信息科技有限公司 | Bank credit risk early warning system and method thereof |
| CN116452316A (en) * | 2023-02-28 | 2023-07-18 | 中国工商银行股份有限公司 | Post-loan risk monitoring method and device for agricultural loan and electronic equipment |
| CN116681889A (en) * | 2023-05-16 | 2023-09-01 | 南京南瑞信息通信科技有限公司 | Multi-branch image semantic segmentation method and system based on AM and feature fusion |
| CN116777608A (en) * | 2023-08-25 | 2023-09-19 | 北京佳格天地科技有限公司 | Agricultural financial risk supervision system, method and storage medium based on big data |
| CN117252689A (en) * | 2023-11-13 | 2023-12-19 | 北京佳格天地科技有限公司 | Agricultural user credit decision support method and system based on big data |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180330435A1 (en) * | 2017-05-11 | 2018-11-15 | Harvesting Inc. | Method for monitoring and supporting agricultural entities |
-
2024
- 2024-02-05 CN CN202410161725.2A patent/CN117710096B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116452316A (en) * | 2023-02-28 | 2023-07-18 | 中国工商银行股份有限公司 | Post-loan risk monitoring method and device for agricultural loan and electronic equipment |
| CN116310822A (en) * | 2023-03-09 | 2023-06-23 | 中国工商银行股份有限公司 | Loan request processing method and device for cultivated land and electronic equipment |
| CN116308754A (en) * | 2023-03-22 | 2023-06-23 | 广州信瑞泰信息科技有限公司 | Bank credit risk early warning system and method thereof |
| CN116681889A (en) * | 2023-05-16 | 2023-09-01 | 南京南瑞信息通信科技有限公司 | Multi-branch image semantic segmentation method and system based on AM and feature fusion |
| CN116777608A (en) * | 2023-08-25 | 2023-09-19 | 北京佳格天地科技有限公司 | Agricultural financial risk supervision system, method and storage medium based on big data |
| CN117252689A (en) * | 2023-11-13 | 2023-12-19 | 北京佳格天地科技有限公司 | Agricultural user credit decision support method and system based on big data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117710096A (en) | 2024-03-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108629413B (en) | Neural network model training and transaction behavior risk identification method and device | |
| Zhang et al. | Using gaussian distribution to construct fitness functions in genetic programming for multiclass object classification | |
| US11263644B2 (en) | Systems and methods for detecting unauthorized or suspicious financial activity | |
| Zhang et al. | A GANs-based deep learning framework for automatic subsurface object recognition from ground penetrating radar data | |
| US11436443B2 (en) | Testing machine learning (ML) models for robustness and accuracy using generative deep learning | |
| CN110634060A (en) | User credit risk assessment method, system, device and storage medium | |
| CN104063686A (en) | System and method for performing interactive diagnosis on crop leaf segment disease images | |
| CN113609889A (en) | High-resolution remote sensing image vegetation extraction method based on sensitive feature focusing perception | |
| CN116777608A (en) | Agricultural financial risk supervision system, method and storage medium based on big data | |
| CN108399423A (en) | A kind of multidate-Combining Multiple Classifiers of classification of remote-sensing images | |
| CN111369152A (en) | Agricultural land value evaluation optimization method, device and equipment and readable storage medium | |
| CN115018513A (en) | Data inspection method, device, equipment and storage medium | |
| Yadav et al. | Geological information extraction from satellite imagery using machine learning | |
| CN114298815A (en) | Listed company financial abnormity identification method and system | |
| CN117710096B (en) | Post-loan risk monitoring method and system for agricultural loan based on land parcel information | |
| CN113159874A (en) | Method and device for detecting value-added tax invoice and readable storage medium | |
| CN117368862A (en) | High-efficiency weather radar data quality evaluation system | |
| CN117196808A (en) | Mobility risk prediction method and related device for peer business | |
| CN116824488A (en) | Target detection method based on transfer learning | |
| CN116011822A (en) | Store cashing management method and device, computer equipment and storage medium | |
| CN115641575A (en) | Leafhopper agricultural pest detection method based on sparse candidate frame | |
| CN113269433B (en) | Tax risk prediction method, apparatus, medium and computer program product | |
| CN114898181A (en) | Hidden danger violation identification method and device for explosion-related video | |
| CN114663899A (en) | Financial bill processing method, device, equipment and medium | |
| Kaur et al. | Deep learning: survey of environmental and camera impacts on internet of things images |
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 |