Farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence

Abstract

The application relates to the technical field of big data analysis and artificial intelligence application, and discloses a farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence, which comprises a farmer information acquisition module, a historical data clustering module, a risk parameter screening module, a trend prediction module, a dynamic correction module and an evaluation scheme generation module. The system acquires target farmer main body information and entrepreneurship project characteristic information, calls historical financing data of similar farmers and carries out clustering analysis, extracts key variables to generate a historical risk parameter set. A risk parameter sequence is formed through optimization and screening, a risk increase rate and a risk decrease rate are predicted, the prediction result is dynamically corrected in combination with entrepreneurship project characteristic differences, and finally a risk assessment prompt and financing operation guidance are generated. The application solves the problems of insufficient data utilization, large correction deviation and weak decision support of the traditional method, and significantly improves the accuracy and practicality of farmer entrepreneurship financing risk assessment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data analysis and artificial intelligence application, in particular to a farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence. BACKGROUND

[0002] Traditional risk assessment methods mainly rely on static financial indicators or manual experience judgment, which have problems such as single data dimension, poor timeliness, and strong subjectivity. Although some existing technologies try to introduce credit scoring models, they fail to fully combine the dynamic characteristics of farmer entrepreneurship projects, such as industry type, operating scale changes, and other key factors, resulting in a large deviation between the assessment results and the actual risks.

[0003] Currently, big data and artificial intelligence technologies provide new solutions for financing risk assessment. For example, clustering analysis can identify risk patterns in historical data, and machine learning models can predict risk trends. However, existing systems still face the following limitations in the context of farmer entrepreneurship: first, historical data is not fully utilized, and there is no differentiated clustering for similar farmer characteristics; second, the risk parameter screening lacks a dynamic optimization mechanism, making it difficult to adapt to changes in the operating environment of farmers; third, the correction process does not consider the fine-grained differences between entrepreneurship project characteristics and standard characteristics, leading to cumulative prediction result bias. In addition, the evaluation scheme generated by traditional systems lacks operational guidance and cannot provide specific decision support for financial institutions.

[0004] Therefore, an evaluation system that can integrate multi-dimensional data, dynamically optimize parameters, and accurately correct risk trends is needed to solve the accuracy, timeliness, and practicality problems in farmer entrepreneurship financing risk assessment. The present application realizes the whole process optimization from data collection to decision generation through modular design, providing financial institutions with a scientific and reliable risk management tool. SUMMARY

[0005] The present application aims to provide a farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence to solve the problems raised in the background.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical solution: a farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence, comprising:

[0007] A farmer information collection module is used to obtain the subject information of the target farmer applying for financing and collect the entrepreneurship project characteristic information of the target farmer, and according to the subject information of the target farmer, historical financing data of similar farmers in the user's historical time is retrieved;

[0008] a historical data clustering module, configured to cluster historical risk influencing factors in the historical financing data, obtain a plurality of clustered risk feature groups, extract a set of historical key variable information and a plurality of historical financing operation sets under the plurality of clustered risk feature groups, and process to generate a plurality of historical risk parameter sets;

[0009] a risk parameter screening module, configured to respectively optimize screen the plurality of historical risk parameter sets, obtain a plurality of screened risk parameter groups, and arrange in time sequence to form a plurality of historical risk parameter sequences;

[0010] a trend prediction module, configured to perform financing risk trend prediction according to the plurality of historical risk parameter sequences, and obtain a risk increase rate and a risk decrease rate;

[0011] a dynamic correction module, configured to match the target farmer's entrepreneurial project feature information with the plurality of clustered risk feature groups, determine a matched risk feature group, and correct the risk increase rate and the risk decrease rate according to differences between the entrepreneurial project feature information and standard feature information of the matched risk feature group, to obtain a corrected increase rate and a corrected decrease rate;

[0012] an evaluation scheme generation module, configured to perform financing risk evaluation decision according to the corrected increase rate and the corrected decrease rate, generate a risk evaluation prompt, and provide financing operation guidance.

[0013] Preferably, target farmer subject information of a current application for financing is obtained, and entrepreneurial project feature information of the target farmer is collected, the target farmer subject information is used to retrieve historical financing data of similar farmers in a historical time of the user, including:

[0014] identity attribute information and credit record information of the target farmer are obtained, and industry type and operating scale information of the entrepreneurial project of the target farmer are collected as entrepreneurial project feature information;

[0015] the target farmer subject information is used to perform associated retrieval in a historical financing database, and historical financing data of similar farmers is obtained.

[0016] Preferably, historical risk influencing factors in the historical financing data are clustered to obtain a plurality of clustered risk feature groups, a set of historical key variable information and a plurality of historical financing operation sets under the plurality of clustered risk feature groups are extracted, and a plurality of historical risk parameter sets are processed and generated, including:

[0017] a plurality of risk influencing factors in the historical financing data are extracted, feature clustering processing is performed, and a plurality of clustered risk feature groups are obtained;

[0018] Collecting the operating stability indicators of farmers in the historical financing data under the multiple clustered risk feature groups to form a historical key variable information set, and extracting the financing application parameter set and financing processing time set submitted by users under the multiple clustered risk feature groups;

[0019] Classify and generate multiple historical basic parameter sets according to the degree of deviation between the multiple financing application parameter sets and the historical key variable information set;

[0020] According to the comparison results between the preset time threshold and the multiple financing processing time sets, the multiple historical basic parameter sets are adjusted and calculated to generate multiple historical risk parameter sets.

[0021] Preferably, the plurality of historical risk parameter sets are optimized and screened respectively to obtain a plurality of screened risk parameter groups, which are arranged in chronological order to form a plurality of historical risk parameter sequences, including:

[0022] Selecting a first benchmark risk parameter in a first historical risk parameter set among the plurality of historical risk parameter sets;

[0023] Assigning selection probabilities based on the degree of difference between other parameters in the first historical risk parameter set and the first benchmark risk parameter to form a first basic probability distribution, wherein the degree of difference is negatively correlated with the selection probability;

[0024] Optimizing and screening the first set of historical risk parameters based on the first basic probability distribution to obtain a first screening risk parameter group;

[0025] Arrange the time identification information of multiple parameters in the first screening risk parameter group in sequence to form the first historical risk parameter sequence;

[0026] Repeat the optimization screening and time series arrangement operations on the remaining multiple historical risk parameter sets to form multiple historical risk parameter sequences.

[0027] Preferably, the first historical risk parameter set is optimized and screened based on the first basic probability distribution to obtain a first screened risk parameter group, including:

[0028] Randomly selecting a preset number of risk parameters from the first historical risk parameter set to form a first screening risk parameter group;

[0029] Allocating selection probabilities based on the degree of difference between the parameters within the first screening risk parameter group and the first benchmark risk parameter to form a first screening probability distribution;

[0030] Calculating the similarity between the first screening probability distribution and the first basic probability distribution as the first screening fitness;

[0031] Randomly select a preset number of risk parameters in the first set of historical risk parameters again to form a second set of screening risk parameters, and calculate a second screening fitness;

[0032] The optimization screening operation is continuously performed until a convergence condition is reached, and the screening risk parameter group with the highest fitness is output as the first screening risk parameter group.

[0033] Preferably, the financing risk trend prediction is performed according to the plurality of historical risk parameter sequences to obtain a risk increase rate and a risk decrease rate, including:

[0034] Sample financing data of a plurality of farmers is collected, a set of sample risk parameter sequences is extracted, and a set of sample increase rates and a set of sample decrease rates are obtained according to parameter change characteristics in each sample risk parameter sequence;

[0035] A risk trend prediction model is constructed by taking the set of sample risk parameter sequences as input features and taking the set of sample increase rates and the set of sample decrease rates as output targets;

[0036] The risk trend prediction model is used to perform risk trend classification prediction on the plurality of historical risk parameter sequences to obtain a plurality of characteristic increase rates and a plurality of characteristic decrease rates;

[0037] The similarity of the target farmer's entrepreneurial project characteristic information to the plurality of clustered risk characteristic groups is analyzed, and the plurality of characteristic increase rates and the plurality of characteristic decrease rates are weighted calculated according to the numerical values of the plurality of similarity degrees to obtain a risk increase rate and a risk decrease rate.

[0038] Preferably, the target farmer's entrepreneurial project characteristic information is matched with the plurality of clustered risk characteristic groups to determine a matching risk characteristic group, and the risk increase rate and the risk decrease rate are corrected according to the difference between the entrepreneurial project characteristic information and the standard characteristic information of the matching risk characteristic group to obtain a corrected increase rate and a corrected decrease rate, including:

[0039] The clustered risk characteristic group with the highest similarity is selected as the matching risk characteristic group, and the standard characteristic information of the matching risk characteristic group is obtained;

[0040] A risk correction coefficient is set according to the difference between the entrepreneurial project characteristic information and the standard characteristic information of the matching risk characteristic group;

[0041] The risk correction coefficient is used to correct and calculate the risk increase rate and the risk decrease rate to obtain a corrected increase rate and a corrected decrease rate.

[0042] Preferably, according to the modified up-rate and the modified down-rate, a financing risk assessment decision is made, a risk assessment prompt is generated, and a financing operation guide is provided, including:

[0043] A sample modified up-rate set and a sample modified down-rate set are collected, and according to the numerical size of each sample modified up-rate and sample modified down-rate, the sample evaluation prompt content is set, and a sample evaluation prompt set is formed, wherein each sample evaluation prompt contains a parameter adjustment amplitude, and the numerical size of the sample modified up-rate and the sample modified down-rate is negatively correlated with the parameter adjustment amplitude;

[0044] The sample modified up-rate set and the sample modified down-rate set are taken as input variables, and the sample evaluation prompt set is taken as output results to construct a risk assessment decision model;

[0045] The risk assessment decision model is used to make an evaluation decision on the modified up-rate and the modified down-rate, and generate a risk assessment prompt.

[0046] Preferably, sample financing data of multiple farmers are collected, a sample risk parameter sequence set is extracted, and a sample up-rate set and a sample down-rate set are obtained according to the parameter change characteristics in each sample risk parameter sequence, including:

[0047] Sample data of farmers with complete financing records in the past three years are screened, and a risk assessment parameter time sequence of each sample is extracted;

[0048] The increase and decrease of the parameters at adjacent time points in each sample risk parameter time sequence are analyzed, and the risk up-rate and the risk down-rate of each sample are calculated;

[0049] The risk parameter time sequences of all samples are summarized into a sample risk parameter sequence set, and the risk up-rate and the risk down-rate of all samples are summarized into a sample up-rate set and a sample down-rate set, respectively.

[0050] Preferably, according to the difference value between the start-up project characteristic information and the standard characteristic information of the matching risk characteristic group, a risk modification coefficient is set, including:

[0051] The difference value between the start-up project characteristic information and the standard characteristic information in the industry type dimension is calculated, and is recorded as a first difference value;

[0052] The difference value between the start-up project characteristic information and the standard characteristic information in the business scale dimension is calculated, and is recorded as a second difference value;

[0053] The first difference value and the second difference value are weighted and summed to obtain a comprehensive difference value;

[0054] According to the preset difference value and the corresponding relationship table of the correction coefficient, the corresponding value of the comprehensive difference value is searched as a risk correction coefficient.

[0055] Compared with the prior art, the present application has the following advantages:

[0056] The farmer entrepreneurship financing risk assessment system provided by the present application realizes intelligent dynamic optimization of the whole process of risk assessment through the cooperative work of multiple modules. The system first uses the farmer information acquisition module to obtain the detailed subject information of the target farmer and the characteristics of the entrepreneurship project, ensuring the comprehensiveness of the data dimension. The historical data clustering module groups the historical financing data of similar farmers, extracts key variables to generate a risk parameter set, and significantly improves the utilization efficiency of historical data. The risk parameter screening module uses a probability distribution optimization method to dynamically screen time sequence parameters, effectively reducing the interference of noise data on the prediction results.

[0057] The trend prediction module analyzes the change law of the risk parameter sequence based on a machine learning model, accurately predicts the risk increase rate and decrease rate, and provides a reliable basis for subsequent correction. The dynamic correction module calculates the correction coefficient to adjust the predicted value by matching the differences between the characteristics of the entrepreneurship project and the clustered standard characteristics, solving the evaluation distortion problem caused by feature deviation in traditional methods. Finally, the evaluation scheme generation module generates specific risk assessment prompt operation guidelines based on the correction results, helping financial institutions to develop differentiated financing strategies and improve risk control capabilities.

[0058] The system has shown significant advantages in practical application. Through the combination of clustering analysis and dynamic correction, the system can adapt to the needs of farmers of different industry types and operating scales, and the evaluation results are more targeted. Optimizing and screening time sequence prediction technology enhances the anti-interference ability of the model, making the risk assessment more stable and reliable. In addition, the financing operation guidelines generated by the system are directly related to the risk parameter adjustment range, providing a practical decision support for financial institutions, reducing the financing threshold for farmers, and promoting the vitality of the rural economy. The present application has outstanding effects in improving evaluation accuracy, enhancing practicality, and optimizing decision-making efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 The working principle diagram of the farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence described in the present application;

[0060] Figure 2 The design diagram of the historical data clustering module;

[0061] Figure 3 The design diagram of the trend prediction module;

[0062] Figure 4 The design diagram of the risk correction coefficient. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work are within the protection scope of the present application.

[0064] Please refer to Figures 1-4 The present application relates to a farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence, and the specific implementation is as follows:

[0065] The farmer information acquisition module acquires the target farmer subject information of the current application for financing, and acquires the entrepreneurship project characteristic information of the target farmer, according to the target farmer subject information, the historical financing data of the same type of farmers in the historical time of the user is called.

[0066] The historical data clustering module clusters the historical risk influencing factors in the historical financing data, obtains a plurality of clustered risk characteristic groups, extracts a historical key variable information set and a plurality of historical financing operation sets under the plurality of clustered risk characteristic groups, and processes to generate a plurality of historical risk parameter sets.

[0067] The risk parameter screening module respectively optimizes and screens the plurality of historical risk parameter sets, obtains a plurality of screened risk parameter groups, and arranges them in time sequence to form a plurality of historical risk parameter sequences.

[0068] The trend prediction module predicts the financing risk trend according to the plurality of historical risk parameter sequences, and obtains the risk rising rate and the risk falling rate.

[0069] The dynamic correction module matches the entrepreneurship project characteristic information of the target farmer with the plurality of clustered risk characteristic groups, determines the matching risk characteristic group, and corrects the risk rising rate and the risk falling rate according to the difference between the entrepreneurship project characteristic information and the standard characteristic information of the matching risk characteristic group, to obtain the corrected rising rate and the corrected falling rate.

[0070] The evaluation scheme generation module makes financing risk assessment decision according to the corrected rising rate and the corrected falling rate, generates risk assessment prompt, and provides financing operation guide. Embodiment 1:

[0071] This embodiment details the specific implementation of the farmer information acquisition module:

[0072] During the system's operation, the first step is to obtain the target farmer's main information. In this process, the collection of identity attribute information covers multiple dimensions. For example, the acquisition of name information needs to ensure that it is completely consistent with the information on the farmer's ID card to ensure the accuracy of identity recognition; the ID number, as a unique identity, will be format-verified and uniquely checked to avoid duplication or errors; the collection of age information not only records the specific value, but also conducts correlation analysis combined with the labor capacity of the farmer and the characteristics of the entrepreneurial project; the information of the place of residence is refined to the specific township and village, which is of great significance for subsequent analysis of the economic development level and policy support of the area where the farmer lives.

[0073] In terms of credit record information collection, the system will interface with multiple credit information platforms. Among them, the bank credit system is an important data source for obtaining information such as whether the farmer has overdue payment records, credit card overdraft, loan default times, etc. In addition, the credit records of the farmer in other financial institutions such as rural credit cooperatives and transaction credit data on e-commerce platforms, such as whether there are default orders, refund rates, etc. These multi-dimensional credit record information can fully reflect the credit status of the farmer, providing an important basis for subsequent financing risk assessment.

[0074] For the collection of target farmer entrepreneurial project characteristic information, the determination of industry type needs to understand the specific business content of the farmer's entrepreneurial project. For example, agricultural planting projects need to further clarify the crop species, whether it is an economic crop, the degree of advanced planting technology, etc.; agricultural product processing projects need to understand the product type, processing technology, and extension of the industrial chain; rural e-commerce projects need to master the product categories, e-commerce platform operation, market share, etc. These detailed industry type information helps the system to accurately match the target farmer with similar farmers in historical financing data.

[0075] The collection of operating scale information also needs to refine multiple indicators. The measurement of planting area will combine satellite remote sensing data and field research results to ensure the accuracy of the data; the number of employees, including full-time and part-time employees, reflects the labor scale of the project; registered capital information needs to be checked with business registration data to ensure the authenticity of the data; the collection of annual sales will refer to multiple data such as the farmer's financial statements and bank flow to fully reflect the operating status of the project.

[0076] After obtaining the target farmer's subject information and entrepreneurial project characteristic information, the system will enter the historical financing data retrieval link. First, the system will use the target farmer's identity attribute information and credit record information as retrieval conditions to perform associated retrieval in the historical financing database. During the retrieval process, a combination of fuzzy matching and exact matching will be used. For example, for the household registration location information, exact township-level matching will be performed first, and if the matching result is small, the range will be expanded to county-level or city-level for fuzzy matching; for credit record information, a certain credit score threshold will be set, and farmers with credit scores near the threshold will be included in the search range.

[0077] When searching for similar farmers, the system will calculate the similarity between the target farmer and the historical farmers according to the pre-set similarity algorithm. This similarity algorithm will consider multiple factors such as identity attribute information, credit record information, and entrepreneurial project characteristic information. For example, factors such as age and household registration location in identity attribute information will be given a certain weight, credit score and overdue times in credit record information will also have corresponding weights, and industry type and business scale in entrepreneurial project characteristic information will also be given weights. Through weighted calculation, the similarity value of each historical farmer to the target farmer is obtained, and the system will select historical farmers with similarity values higher than the pre-set threshold to obtain their historical financing data.

[0078] The content of historical financing data includes multiple aspects. Historical financing application information covers financing amount, financing period, application time, etc.; approval result information includes whether to pass the approval, approval time, approval amount, etc.; and repayment situation information includes repayment time, repayment amount, whether to repay on time, etc. These data will be completely retrieved and stored by the system for subsequent clustering analysis and risk assessment.

[0079] To ensure the accuracy and completeness of the data, the system will perform multiple checks on the collected and retrieved data. For identity attribute information and credit record information, it will be compared with authoritative databases; for entrepreneurial project characteristic information, farmers will be required to provide relevant proof materials such as business license, land contract, financial statements, etc., and professional personnel will perform audits; for historical financing data, the consistency and completeness of the data will be checked to ensure that there is no missing or incorrect information.

[0080] In terms of data storage, the system will use distributed database technology to store the collected and retrieved data on multiple servers to improve data security and availability. At the same time, data encryption will be performed to prevent data leakage.

[0081] In addition, the system also establishes a data update mechanism. When the information of target farmers changes, such as credit record updates, business project scale expansion, etc., the system will update the relevant information in time; for historical financing data, the latest data will also be obtained regularly from the data source to ensure that the data in the database always remains up-to-date. Embodiment 2:

[0082] This embodiment describes in detail the implementation of the historical data clustering module:

[0083] When processing historical financing data, the system needs to extract factors that have an impact on financing risk from massive data. These historical risk impact factors cover multiple dimensions, including the operating status of farmers themselves, the development trend of the industry they are in, the macroeconomic environment, etc. For example, in terms of the operating status of farmers, factors such as operating years, annual profit fluctuation, asset-liability ratio, fixed asset scale, etc. are extracted; in terms of industry development trend, factors such as industry policy support intensity, market demand change trend, industry competition intensity, etc. are included; macroeconomic environment involves factors such as local GDP growth rate, inflation rate, employment rate, etc. The system will comprehensively sort out these factors to ensure that important risk impact factors are not missed.

[0084] After extracting multiple risk impact factors, the system will perform feature clustering processing on these factors. Feature clustering processing uses an advanced clustering algorithm that can group risk impact factors into different groups according to their similarity. In the clustering process, appropriate clustering parameters need to be set, such as the number of clusters, the calculation method of similarity, etc. The calculation of similarity takes into account the numerical difference and attribute characteristics of multiple factors. For numerical factors, methods such as Euclidean distance are used to calculate the difference; for categorical factors, appropriate classification similarity algorithms are used for calculation. By continuously adjusting the clustering parameters, the risk impact factors within each clustering group have a high degree of similarity, and the differences between different clustering groups are large, so that multiple clustering risk feature groups are obtained.

[0085] After clustering is completed, the system needs to collect the historical key variable information set under each clustering risk feature group. The historical key variable information set is mainly composed of operating stability indicators of farmers. Taking agricultural planting farmers as an example, operating stability indicators may include continuous planting years, yield fluctuation amplitude of main crops, perfection degree of irrigation facilities, ability to cope with natural disasters, etc.; for agricultural product processing farmers, operating stability indicators may involve the advanced degree of processing equipment, stability of raw material supply, diversity of product sales channels, etc. The system will extract these indicators one by one from historical financing data to ensure that the data source of each indicator is reliable and the record is accurate.

[0086] Meanwhile, the system will also extract the financing application parameter sets and financing processing time sets submitted by users under multiple clustering risk feature groups. The financing application parameter set includes the financing amount, expected financing period, acceptable interest rate range, and repayment method information proposed by the farmer when applying for financing; the financing processing time set records the time spent from the submission to the final approval of each financing application. These information is of great significance for subsequent analysis of financing risk parameters.

[0087] The system will classify historical financing data according to the deviation degree of multiple financing application parameter sets and historical key variable information sets, generating multiple historical base parameter sets. The calculation of deviation degree needs to consider the difference between each financing application parameter and the corresponding historical key variable. For example, if the financing amount applied by a farmer is much higher than the asset size and profitability reflected in the historical key variable information, the deviation degree of the financing application parameter and the historical key variable information is large. The system will set a reasonable deviation threshold for each financing application parameter, and divide the historical financing data into different categories according to the size of the deviation degree, each category corresponds to a historical base parameter set.

[0088] After generating the historical base parameter set, the system needs to adjust and calculate multiple historical base parameter sets according to the comparison results of the preset time threshold and multiple financing processing time sets. The setting of the preset time threshold will refer to the average financing processing time in the industry, and make appropriate adjustments combined with different financing types and sizes. The system will compare each financing processing time with the preset time threshold, if the financing processing time exceeds the threshold, it means that there may be some special circumstances in the approval process of the financing, such as the difficulty of farmer's data audit, the complexity of project risk assessment, etc. For these situations, the system will adjust the parameters in the corresponding historical base parameter set, for example, increase the weight of risk related parameters, to reflect the additional risks that may exist in the approval process. The adjustment and calculation process needs to consider multiple factors to ensure that the adjusted historical risk parameter set can more accurately reflect the financing risk situation.

[0089] In the whole process of historical data clustering, data quality control is crucial. The system will strictly verify and clean the extracted historical risk influencing factors, operating stability indicators, financing application parameters, etc. For missing data, reasonable interpolation method will be used for filling; for abnormal data, identification and correction will be carried out to ensure the accuracy and integrity of the data. At the same time, the system will regularly evaluate and optimize the clustering results, according to the changes of new historical financing data and business needs, adjust the clustering parameters and calculation methods, so that the clustering risk feature group and the historical risk parameter set can always accurately reflect the actual situation.

[0090] In addition, the system also establishes an updating mechanism for the historical risk parameter set. When new historical financing data is entered, the system automatically incorporates it into the clustering analysis process, regenerates the clustered risk feature group and the historical risk parameter set, ensuring the timeliness and accuracy of the data. Through this dynamic updating method, the system can continuously learn and adapt to new risk features, improving the accuracy and reliability of financing risk assessment.

[0091] Through the above detailed implementation, the historical data clustering module can effectively cluster the historical financing data, generating a historical risk parameter set that accurately reflects the financing risk features, providing strong support for subsequent risk parameter screening and risk trend prediction. The entire process strictly follows the principles of data processing standardization and scientificity, ensuring that each link accurately reflects the internal characteristics and risk patterns of historical financing data. Example 3:

[0092] This embodiment describes the specific implementation of the risk parameter screening module:

[0093] When performing risk parameter screening, the system first needs to select a first benchmark risk parameter within the first historical risk parameter set. The first historical risk parameter set contains multiple parameters related to financing risk extracted from historical financing data, such as the asset-liability ratio of farmers, operating years, industry risk coefficients, etc. The selection of the benchmark risk parameter needs to consider the importance and representativeness of the parameter in financing risk assessment, and usually selects parameters with high correlation to financing risk in historical data, such as the asset-liability ratio, which can directly reflect the debt repayment ability of farmers and has an important impact on financing risk assessment.

[0094] After selecting the first benchmark risk parameter, the system calculates the difference between other parameters in the first historical risk parameter set and the benchmark risk parameter, and assigns selection probabilities according to the difference, forming the first basic probability distribution. The calculation of the difference considers the type and value characteristics of the parameters. For numerical parameters, the difference is measured by calculating the absolute difference or relative difference between the parameter value and the benchmark parameter value; for categorical parameters, the difference is determined by judging the difference between the parameter category and the benchmark parameter category. The difference and the selection probability are negatively correlated, i.e. the smaller the difference between a parameter and the benchmark risk parameter, the higher the probability of being selected in the screening process, and vice versa. For example, if the benchmark risk parameter is an asset-liability ratio of 25%, another parameter is an asset-liability ratio of 28%, the difference is small, and the selection probability is high; while if a parameter is an asset-liability ratio of 45%, the difference is large, and the selection probability is low.

[0095] According to the first base probability distribution, the system starts to optimize and screen the first historical risk parameter set. In specific operation, the system will randomly select a preset number of risk parameters from the first historical risk parameter set to form the first screening risk parameter group. The preset number needs to consider the size of the parameter set and the accuracy requirement of risk assessment, and is usually set to a number that can cover the main risk characteristics, such as randomly selecting 10 parameters from a set containing 20 parameters.

[0096] After forming the first screening risk parameter group, the system will calculate the difference degree of the parameters in the group and the first benchmark risk parameter, and redistribute the selection probability according to the difference degree to form the first screening probability distribution. Then, the system will calculate the similarity degree of the first screening probability distribution and the first base probability distribution as the first screening fitness. The calculation of similarity degree uses a probability distribution similarity algorithm, which can measure the closeness between two probability distributions.

[0097] The system will again randomly select a preset number of risk parameters from the first historical risk parameter set to form the second screening risk parameter group, and calculate the second screening fitness in the same way. This random selection and fitness calculation operation is repeated to continue the optimization and screening.

[0098] During the optimization and screening process, the system will set a convergence condition. When the screening fitness no longer significantly improves or reaches a preset number of iterations, it is considered that the screening process has converged. At this time, the system will output the screening risk parameter group with the highest fitness as the first screening risk parameter group. This process is similar to finding the combination that best matches the base probability distribution among many possible parameter combinations, to ensure that the selected parameter group can best reflect the characteristics of the historical risk parameter set.

[0099] After obtaining the first screening risk parameter group, the system will arrange the multiple parameters in the group in chronological order according to their time identifier information to form the first historical risk parameter sequence. The time identifier information records the time point corresponding to each parameter, such as the financing application time, approval time, etc. By arranging the parameters in chronological order, the trend of risk parameters changing over time can be reflected.

[0100] After completing the screening and sorting of the first historical risk parameter set, the system will repeat the above optimization and screening and time sequence arrangement operations for the remaining multiple historical risk parameter sets. For each historical risk parameter set, a benchmark risk parameter needs to be selected first, the difference degree and the base probability distribution are calculated, then random selection and fitness calculation are performed, until the screening risk parameter group is obtained, and finally the historical risk parameter sequence is arranged in chronological order.

[0101] During the entire risk parameter screening process, the processing and calculation of data need to maintain a high degree of accuracy and consistency. The system will conduct strict data verification on each historical risk parameter set to ensure the accuracy of the parameter values and time identification information. For missing time identification information, the system will complete it by associating other data records; for abnormal parameter values, it will be checked and corrected.

[0102] At the same time, the system will dynamically adjust the preset number and convergence conditions according to the characteristics of different historical risk parameter sets. For example, for a set with a large number of parameters, the preset number can be increased appropriately to ensure that the selected parameter group contains sufficient risk characteristics; for a set with high data quality, the convergence condition can be appropriately improved to obtain more accurate screening results.

[0103] In addition, the system also establishes a verification mechanism for the screening results. For the historical risk parameter sequence obtained by screening, it is verified by comparing with the actual financing risk results to verify its effectiveness and accuracy. If it is found that the screening result deviates greatly from the actual situation, the screening process will be reviewed again, and the relevant parameters and algorithms will be adjusted to ensure that the selected risk parameters can accurately reflect the trend of financing risk. Embodiment 4:

[0104] In the farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence, embodiment 4 is as follows for the specific implementation of the trend prediction module:

[0105] When the system predicts the trend of financing risk, it needs to collect sample financing data of multiple farmers. Taking the historical data of a county rural financial institution as an example, the system selects samples of farmers with complete financing records in the past three years, such as selecting 1000 farmers covering different industries such as planting, breeding, and rural e-commerce. These sample data need to include full-process information from financing application to repayment completion, including farmers' basic identity information, entrepreneurship project characteristics, historical risk parameters, and corresponding risk assessment results, etc.

[0106] The system extracts the risk assessment parameter time series from each sample. Taking a planting farmer, Mr. Zhang, as an example, his risk parameter time series may include asset-liability ratio 35% in January 2022, operating period 5 years, industry risk coefficient 0.6, asset-liability ratio 38% in July 2022, operating period 5.5 years, industry risk coefficient 0.7 (affected by natural disasters that year), asset-liability ratio 36% in January 2023, operating period 6 years, industry risk coefficient 0.5, etc. Each time point parameter corresponds to a specific financing application or post-loan management node. The system aggregates the parameter time series of 1000 samples into a sample risk parameter sequence set, each sequence containing at least 3 time node parameter data.

[0107] The system analyzes the increase and decrease of the parameters at adjacent time points in each sample risk parameter time series, and calculates the sample rising rate and falling rate. Taking Zhang's asset-liability ratio as an example, it rose from 35% to 38% from January to July 2022, with a rising rate of (38%-35%) / 35% ≈ 8.57%; it fell from 38% to 36% from July 2022 to January 2023, with a falling rate of (38%-36%) / 38% ≈ 5.26%. The rising rate and falling rate are calculated for all risk parameters (such as asset-liability ratio, industry risk coefficient, etc.) of each sample, and the average value is taken as the risk rising rate and falling rate of the sample. For example, the comprehensive rising rate of Zhang is the arithmetic mean of the rising rates of each parameter, and the comprehensive falling rate is the same. The rising rates and falling rates of the 1000 samples are respectively summarized into a sample rising rate set (containing 1000 rising rate values) and a sample falling rate set.

[0108] The system takes the sample risk parameter sequence set as the input feature, and takes the sample rising rate set and the sample falling rate set as the output target, and constructs a risk trend prediction model. The model uses recurrent neural network (RNN) or long short-term memory network (LSTM) in machine learning, because it is suitable for processing time series data. Taking the LSTM model as an example, the input layer receives the risk parameters at each time node (such as asset-liability ratio, operating years, etc.), the hidden layer captures the trend of parameter changes over time through memory cells, and the output layer predicts the corresponding rising rate and falling rate. During model training, 800 samples are used as the training set and 200 samples are used as the validation set. By adjusting the number of network layers, the number of neurons and other hyperparameters, the prediction error of the model on the validation set is minimized.

[0109] After the model is constructed, the system uses the model to classify and predict the risk trend of multiple historical risk parameter sequences. For example, for a historical risk parameter sequence containing asset-liability ratio 40% and operating years 3 years in January 2023, asset-liability ratio 42% and operating years 3.5 years in July 2023, asset-liability ratio 45% and operating years 4 years in January 2024, the model will predict that the feature rising rate is 10% and the feature falling rate is 3%, reflecting that the risk of this type of farmer is on the rise.

[0110] Meanwhile, the system needs to analyze the similarity between the target farmer's entrepreneurial project characteristic information and multiple clustering risk characteristic groups. Taking target farmer Li (engaged in fruit planting) as an example, his entrepreneurial project characteristics are industry type "economic crop planting" and operating scale "planting area 50 mu, annual turnover 800,000 yuan". The system matches these characteristics with the risk characteristic groups formed by historical clustering (such as "economic crop planting-small scale", "grain planting-large scale", etc.), and calculates the similarity of each characteristic group. The calculation of similarity is based on the matching degree of industry type (such as "economic crop planting" and the industry type of a certain characteristic group completely matching, scoring 1 point, and partially matching scoring 0.5 points) and the numerical difference of operating scale (such as the difference between the normalized processing area and the turnover and the standard value of the characteristic group). Finally, the comprehensive similarity score of each characteristic group is obtained. Assuming that Li's similarity with the "economic crop planting-medium scale" characteristic group is the highest, which is 0.85, and the similarity with other characteristic groups is 0.6, 0.4, etc.

[0111] According to the numerical size of multiple similarity degrees, the system performs weighted calculation on the multiple characteristic rising rates and characteristic falling rates predicted by the model. The weighting formula is: risk rising rate = Σ (characteristic rising rate x corresponding characteristic group similarity) / Σ similarity, and the risk falling rate is the same. For example, the model predicts the characteristic rising rates of the three characteristic groups corresponding to Li as 8%, 10%, and 12%, respectively, and the similarity degrees are 0.85, 0.1, and 0.05, respectively. Then the risk rising rate = (8% x 0.85 + 10% x 0.1 + 12% x 0.05) / (0.85 + 0.1 + 0.05) = (6.8% + 1% + 0.6%) / 1 = 8.4%, and the risk falling rate is calculated according to the same logic.

[0112] During the entire implementation process, the screening and preprocessing of sample data is crucial. The system will exclude samples with incomplete financing records (such as missing post-loan management data), and correct or mark abnormal parameter values (such as asset-liability ratio exceeding 100% without reasonable explanation). For samples with insufficient time series length (such as containing only 1 time node), the intermediate node data is supplemented by interpolation method to ensure the consistency of model input.

[0113] In addition, the updating mechanism of the model also needs to run continuously. When more than 500 new sample data are added, the system will retrain the LSTM model and adjust the parameter weights to adapt to the new risk trend characteristics. For example, if the risk parameter change trend of the rural e-commerce industry in the new data is different from the past, the model will capture this change through retraining to avoid prediction bias. Embodiment 5

[0114] In the farmer entrepreneurial financing risk assessment system based on big data and artificial intelligence, the specific implementation of the dynamic correction module and the evaluation scheme generation module is as follows:

[0115] In the implementation of the dynamic correction module, the system needs to match the target farmer's entrepreneurial project characteristic information with multiple clustering risk characteristic groups to determine the matching risk characteristic group. Taking the target farmer Wang as an example, his entrepreneurial project is "100 tons of agricultural product processing enterprise", the industry type belongs to agricultural product processing, and the operation scale is represented by the factory area of 1000 square meters, the number of employees of 20 people, and the annual turnover of 5 million yuan. The system will compare these characteristic information with multiple risk characteristic groups formed by clustering historical data, such as "agricultural product processing-small enterprise", "agricultural product processing-medium enterprise", "planting-large scale" and other characteristic groups.

[0116] In the matching process, the system will calculate the similarity degree of the entrepreneurial project characteristic information and each clustering risk characteristic group. For the industry type, the category matching method is adopted, if the industry type of the target farmer is completely consistent with a certain characteristic group, the industry dimension score is 1; if it belongs to the related industry (such as agricultural product processing and agricultural product planting), the score is 0.5. For the operation scale, the specific numerical value (such as factory area, turnover, etc.) of the target farmer needs to be calculated with the standard characteristic information of the characteristic group. For example, the standard characteristic information of "agricultural product processing-small enterprise" characteristic group is factory area 500-1500 square meters, number of employees 10-30 people, annual turnover 300-800 million yuan, Wang's factory area 1000 square meters is within the interval, the score is 1; the number of employees is 20, which is also within the interval, the score is 1; the annual turnover of 500 million yuan is also within the interval, the score is 1. Combining the scores of industry type and operation scale, the system calculates the similarity degree of Wang and "agricultural product processing-small enterprise" characteristic group as (1+1+1) / 3x1(industry weight)=1, and the similarity degree with other characteristic groups is lower than the value, so the characteristic group is selected as the matching risk characteristic group, and the standard characteristic information of the factory area, the number of employees, the annual turnover interval range and the corresponding risk parameter standard value are obtained.

[0117] The system needs to correct the risk increase rate and decrease rate according to the difference between the entrepreneurial project characteristic information and the standard characteristic information of the matching risk characteristic group. Take Wang as an example, although the operating scale value is within the standard interval, the system will further calculate the difference between the specific value and the median value in the standard interval. The median value of the factory area of the "agricultural product processing-small enterprise" characteristic group is 1000 square meters, and the factory area of Wang is exactly the median value, the first difference value is 0; the median value of the number of employees is 20, and the number of employees of Wang is also 20, the second difference value is 0; the median value of the annual sales of the standard interval is 5.5 million yuan, and the annual sales of Wang is 5 million yuan, the difference with the median value is 0.5 million yuan, and the second difference value is obtained by normalizing processing (such as dividing the difference value by the interval span), which is-0.2 (assuming the interval span is 5 million yuan, (500-550) / 500=-0.1, the specific calculation is adjusted according to the actual data range). The first difference value (industry dimension) and the second difference value (operating scale dimension) are weighted and summed according to the preset weight (such as industry dimension weight 0.4, operating scale weight 0.6), and the comprehensive difference value is 0x0.4+(-0.2) x0.6=-0.12.

[0118] The system presets the corresponding relationship table of difference value and correction coefficient, for example, when the comprehensive difference value is in the interval of-0.15 to-0.1, the risk correction coefficient is 0.95; when it is in the interval of-0.1 to 0, the correction coefficient is 1. According to the comprehensive difference value-0.12 of Wang, the risk correction coefficient is determined to be 0.95 according to the table. Assuming that the risk increase rate obtained by the trend prediction module is 10%, and the decrease rate is 5%, the risk correction coefficient is used for correction calculation, the corrected increase rate=10% x0.95=9.5%, and the corrected decrease rate=5% x0.95=4.75%, that is, through the difference adjustment of entrepreneurial project characteristics and standard characteristics, the expected range of risk increase is reduced.

[0119] In the evaluation scheme generation module, the system first collects a set of sample correction up-rate and a set of sample correction down-rate. Taking a historical sample of 1000 households as an example, each sample obtains a corresponding correction up-rate and correction down-rate after dynamic correction, such as a correction up-rate of 8% and a correction down-rate of 6% for sample A, a correction up-rate of 12% and a correction down-rate of 3% for sample B, etc. The system sets the sample evaluation prompt content according to the numerical size of each sample correction up-rate and correction down-rate, forming a set of sample evaluation prompts. For example, when the correction up-rate is ≤5% and the correction down-rate is ≥8%, the evaluation prompt is “low risk, can suggest full approval of financing application”; when the correction up-rate is between 5%-10% and the correction down-rate is between 3%-8%, the prompt is “moderate risk, suggest adjusting the financing amount and shortening the term”; when the correction up-rate is >10%, the prompt is “high risk, suggest rejecting financing or requiring a mortgage guarantee”. Each sample evaluation prompt contains a parameter adjustment amplitude, and the numerical size of the correction up-rate and the correction down-rate is negatively correlated with the parameter adjustment amplitude, that is, the higher the correction up-rate, the greater the recommended financing amount adjustment amplitude (such as reducing the financing amount).

[0120] The system takes the set of sample correction up-rate and the set of sample correction down-rate as input variables, and takes the set of sample evaluation prompts as output results, to construct a risk evaluation decision model. The model can use classification algorithms such as support vector machine (SVM) or random forest, to learn the mapping relationship between correction rate and evaluation prompt through training sample data. For example, when using a random forest model, the correction up-rate and the correction down-rate are used as feature inputs, and the categories of the evaluation prompt (such as “low risk”, “moderate risk”, “high risk”) are used as outputs, to perform ensemble learning through the construction of multiple decision trees, to improve the accuracy of classification.

[0121] Through the risk evaluation decision model, the system evaluates and decides the correction up-rate and the correction down-rate of the target farmer, to generate a risk evaluation prompt. Taking the correction up-rate of 9.5% and the correction down-rate of 4.75% of Wang as an example, the model judges that it belongs to the “moderate risk” category, and generates a risk evaluation prompt of “the farmer's entrepreneurial project has moderate financing risk, it is suggested to adjust the applied financing amount from 2 million yuan to 1.5 million yuan, and to require a fixed asset mortgage”. At the same time, the model provides specific financing operation guidelines, such as recommending the use of quarterly interest payment and end-of-term principal repayment to reduce post-loan management risk, and recommending an increase in post-loan inspection frequency to once every quarter to closely monitor the operating status.

[0122] In the implementation process, the integrity and accuracy of the sample data directly affect the effectiveness of the model. The system will update the sample library regularly to include new financing cases, ensuring that the model can adapt to market environment and policy changes. For example, when the policy support for the agricultural product processing industry is strengthened, the distribution of the correction rate of new samples may change. The system re-trains the model by updating the sample data and adjusts the threshold and content of the evaluation prompt.

[0123] In addition, the difference value calculation and correction coefficient table in the dynamic correction module will be adjusted according to the development of the industry. For example, if the average operating scale of the agricultural product processing industry expands in a certain year, the standard feature information of "small enterprises" may be adjusted to factory area of 800-1800 square meters. The system will update the standard feature information and difference value calculation method synchronously to ensure that the correction process conforms to the current industry actual situation.

[0124] Through the above implementation mode, the dynamic correction module and the evaluation scheme generation module realize the whole process quantification processing from feature matching, difference correction to risk decision-making, combining historical sample data and real-time feature analysis to provide scientific risk evaluation prompts and operation guidelines for farmer entrepreneurship financing, which not only guarantees the risk control needs of financial institutions, but also provides targeted financing suggestions for farmers, and improves the accuracy and efficiency of rural financial services.

[0125] It should be noted that in this paper, relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the term "include" "contain" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0126] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence, characterized by: The system comprises: A farmer information collection module is used to obtain the subject information of the target farmer currently applying for financing, as well as the characteristic information of the target farmer's entrepreneurial project. Based on the target farmer's subject information, the module retrieves the historical financing data of similar farmers within the user's historical time period; A historical data clustering module, configured to cluster historical risk influencing factors within the historical financing data to obtain a plurality of clustered risk feature groups, extract historical key variable information sets and a plurality of historical financing operation sets from the plurality of clustered risk feature groups, and process and generate a plurality of historical risk parameter sets; a risk parameter screening module, wherein the risk parameter screening module is used to optimize and screen the multiple historical risk parameter sets respectively to obtain multiple screened risk parameter groups, and arrange them in chronological order to form multiple historical risk parameter sequences; A trend prediction module, configured to predict financing risk trends based on the multiple historical risk parameter sequences to obtain risk rise and fall rates; a dynamic correction module, the dynamic correction module being used to match the entrepreneurial project characteristic information of the target farmer with the multiple clustered risk characteristic groups, determine a matching risk characteristic group, and correct the risk increase rate and decrease rate based on the difference between the entrepreneurial project characteristic information and the standard characteristic information of the matching risk characteristic group to obtain a corrected increase rate and a corrected decrease rate; An evaluation scheme generation module is used to make financing risk assessment decisions based on the modified increase rate and the modified decrease rate, generate risk assessment prompts, and provide financing operation guidance.

2. The farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence according to claim 1 is characterized in that: Obtain the target farmer's subject information for the current financing application, as well as the entrepreneurial project feature information of the target farmer. Based on the target farmer's subject information, retrieve the user's historical financing data for similar farmers within the historical period, including: Obtaining identity attribute information and credit record information of target farmers, and collecting industry type and business scale information of the target farmers' entrepreneurial projects as entrepreneurial project feature information; Based on the target farmer's main information, a related search is performed in the historical financing database to obtain historical financing data of similar farmers.

3. The farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence according to claim 1 is characterized in that: Clustering the historical risk influencing factors in the historical financing data to obtain multiple clustered risk feature groups, extracting historical key variable information sets and multiple historical financing operation sets under the multiple clustered risk feature groups, and processing to generate multiple historical risk parameter sets, including: Extracting multiple risk influencing factors from the historical financing data, performing feature clustering processing, and obtaining multiple clustered risk feature groups; Collecting the operating stability indicators of farmers in the historical financing data under the multiple clustered risk feature groups to form a historical key variable information set, and extracting the financing application parameter set and financing processing time set submitted by users under the multiple clustered risk feature groups; Classify and generate multiple historical basic parameter sets according to the degree of deviation between the multiple financing application parameter sets and the historical key variable information set; According to the comparison results between the preset time threshold and the multiple financing processing time sets, the multiple historical basic parameter sets are adjusted and calculated to generate multiple historical risk parameter sets.

4. The farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence according to claim 1 is characterized in that: Optimize and screen the multiple historical risk parameter sets respectively to obtain multiple screened risk parameter groups, and arrange them in chronological order to form multiple historical risk parameter sequences, including: Selecting a first benchmark risk parameter in a first historical risk parameter set among the plurality of historical risk parameter sets; Assigning selection probabilities based on the degree of difference between other parameters in the first historical risk parameter set and the first benchmark risk parameter to form a first basic probability distribution, wherein the degree of difference is negatively correlated with the selection probability; Optimizing and screening the first set of historical risk parameters based on the first basic probability distribution to obtain a first screening risk parameter group; Arrange the time identification information of multiple parameters in the first screening risk parameter group in sequence to form the first historical risk parameter sequence; Repeat the optimization screening and time series arrangement operations on the remaining multiple historical risk parameter sets to form multiple historical risk parameter sequences.

5. The farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence according to claim 4 is characterized in that: Based on the first basic probability distribution, the first historical risk parameter set is optimized and screened to obtain a first screened risk parameter group, including: Randomly selecting a preset number of risk parameters from the first historical risk parameter set to form a first screening risk parameter group; Allocating selection probabilities based on the degree of difference between the parameters within the first screening risk parameter group and the first benchmark risk parameter to form a first screening probability distribution; Calculating the similarity between the first screening probability distribution and the first basic probability distribution as the first screening fitness; Again randomly selecting a preset number of risk parameters from the first historical risk parameter set to form a second screening risk parameter group, and calculating a second screening fitness; The optimization and screening operation is continued until the convergence condition is reached, and the screening risk parameter group with the highest fitness is output as the first screening risk parameter group.

6. The farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence according to claim 1 is characterized in that: Financing risk trend forecasting is performed based on the multiple historical risk parameter sequences to obtain risk rise and fall rates, including: Collect sample financing data of multiple farmers, extract a set of sample risk parameter sequences, and obtain a set of sample rising rates and a set of sample falling rates based on the parameter change characteristics within each sample risk parameter sequence; Taking the sample risk parameter sequence set as input features and the sample increase rate set and sample decrease rate set as output targets, a risk trend prediction model is constructed; Using the risk trend prediction model, risk trend classification prediction is performed on the multiple historical risk parameter sequences to obtain multiple characteristic increase rates and multiple characteristic decrease rates; Analyze the similarity between the entrepreneurial project characteristic information of the target farmer and the multiple clustered risk characteristic groups, and perform weighted calculation on the multiple characteristic increase rates and multiple characteristic decrease rates according to the numerical values ​​of the multiple similarities to obtain the risk increase rate and decrease rate.

7. The farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence according to claim 6 is characterized in that: Matching the entrepreneurial project characteristic information of the target farmer with the multiple clustered risk characteristic groups to determine a matching risk characteristic group, and correcting the risk increase rate and decrease rate based on the difference between the entrepreneurial project characteristic information and the standard characteristic information of the matching risk characteristic group to obtain a corrected increase rate and a corrected decrease rate, including: Selecting the clustered risk feature group with the highest similarity as the matching risk feature group, and obtaining standard feature information of the matching risk feature group; Setting a risk correction coefficient according to the difference between the entrepreneurial project characteristic information and the standard characteristic information of the matching risk characteristic group; The risk correction coefficient is used to correct the risk increase rate and decrease rate to obtain a corrected increase rate and a corrected decrease rate.

8. The farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence according to claim 1 is characterized in that: Based on the modified increase rate and the modified decrease rate, a financing risk assessment decision is made, risk assessment prompts are generated, and financing operation guidance is provided, including: Collect a set of sample correction rising rates and a set of sample correction falling rates, and set sample evaluation prompt content according to the numerical value of each sample correction rising rate and sample correction falling rate to form a set of sample evaluation prompts, wherein each sample evaluation prompt includes a parameter adjustment range, and the numerical values ​​of the sample correction rising rate and the sample correction falling rate are negatively correlated with the parameter adjustment range; Constructing a risk assessment decision model using the sample corrected rise rate set and the sample corrected fall rate set as input variables and the sample evaluation prompt set as output results; The risk assessment decision model is used to evaluate and decide on the revised rise rate and the revised fall rate, and generate a risk assessment prompt.

9. The farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence according to claim 6 is characterized in that: Collect sample financing data of multiple farmers, extract the sample risk parameter series set, and obtain the sample increase rate set and sample decrease rate set based on the parameter change characteristics within each sample risk parameter series, including: Screen the sample data of farmers with complete financing records in the past three years and extract the time series of risk assessment parameters for each sample; Analyze the increase and decrease of the parameters at adjacent time points in the risk parameter time series of each sample, and calculate the risk increase and decrease rates of each sample; The risk parameter time series of all samples are summarized into a sample risk parameter series set, and the risk increase rate and decrease rate of all samples are summarized into a sample increase rate set and a sample decrease rate set respectively.

10. The farmer entrepreneurship financing risk assessment system based on big data and artificial intelligence according to claim 7 is characterized in that: The risk correction coefficient is set according to the difference between the entrepreneurial project characteristic information and the standard characteristic information of the matching risk characteristic group, including: Calculate the difference between the entrepreneurial project feature information and the standard feature information in the industry type dimension, and record it as a first difference value; Calculate the difference between the entrepreneurial project characteristic information and the standard characteristic information in the business scale dimension, and record it as a second difference value; Performing a weighted summation on the first difference value and the second difference value to obtain a comprehensive difference value; According to the preset correspondence table between difference values ​​and correction coefficients, the value corresponding to the comprehensive difference value is searched and used as the risk correction coefficient.

Images