A multi-dimensional identity recognition and risk early warning method and system
The multi-dimensional identity recognition and risk warning system solves the problems of single identity recognition and incomplete risk warning in family farm management, realizes accurate identification and automated management, reduces the risk of identity theft, and improves the efficiency and security of fund allocation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHENGDU FINANCIAL HLDG CREDIT INFORMATION CO LTD
- Filing Date
- 2026-03-06
- Publication Date
- 2026-07-21
AI Technical Summary
The current management of family farms relies on a single dimension for identity verification, leading to frequent issues of separation between individuals and businesses and identity theft. Risk warning mechanisms are one-sided and fail to effectively capture operational risks. Traditional manual verification is inefficient and poses security risks, and data is scattered and not effectively correlated.
A multi-dimensional identity recognition and risk warning system was constructed. Multi-dimensional data of family farms were collected from the database of the Ministry of Agriculture and Rural Affairs. The data was cleaned, standardized and correlated. Logistic regression algorithm was used to construct feature engineering. Combined with identity correlation, operational stability and credit performance capability, automated verification and risk warning were achieved. Sensitive information was protected by de-identification display and hierarchical access control.
It enables precise identification of family farms, early risk assessment, reduction of financial security risks, improved management efficiency and security, and reduced financial waste.
Smart Images
Figure CN122434643A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of identification and risk warning technology, and in particular to a multi-dimensional identity recognition and risk warning method and system. Background Technology
[0002] As the core carrier of modern agricultural development, the authenticity of family farms and the stability of their operations directly affect the accurate allocation of funds by financial institutions and the healthy development of the agricultural industry.
[0003] However, existing family farm management technologies have many prominent pain points. Among them, the identification dimension is too narrow, focusing only on single information such as the legal representative's ID card, without linking key data such as social credit code and reserved telephone number. This leads to frequent problems of separation between people and enterprises and identity theft, which allows criminals to take advantage of the situation to obtain financial loans and cause financial losses.
[0004] The risk warning mechanism is one-sided, focusing only on single credit risks such as loan delinquency, and failing to include operational risk indicators such as land area fluctuations and grain production declines. The delayed warnings have resulted in the idle and wasted fiscal funds due to the poor management of operating entities.
[0005] Moreover, relying on the traditional manual verification model is not only inefficient and prone to human error, but also poses a security risk of sensitive information leakage. At the same time, the data within the Ministry of Agriculture and Rural Affairs is scattered across multiple systems without effective correlation and lacks a unified business entity file, making it impossible for financial institutions to fully assess the business status and meet the actual needs of family farms for accurate identification, early risk warning, and efficient and safe management. Summary of the Invention
[0006] The purpose of this invention is to provide a multi-dimensional identity recognition and risk warning method and system to solve the problems mentioned in the background art.
[0007] To achieve the above objectives, the present invention provides the following technical solution: a multi-dimensional identity recognition and risk warning method, S1: data collection and preprocessing, wherein the data collection source is multi-dimensional data of family farms in the database of the Ministry of Agriculture and Rural Affairs, the multi-dimensional data of family farms includes the area of land operated by the family farm, grain output, demonstration farm level, financial support funds, social credit code, registration status, legal representative's ID number, contact number, annual total income, and the area and term of land transfer; the preprocessing includes data cleaning, outlier handling, data standardization, and data association; S2: Feature engineering and model building. The core features extracted by feature engineering include identity correlation, business stability, credit performance capability, and business qualification level. The model is constructed using a logistic regression algorithm. The model training process includes selecting historical data of family farms in the database of the Ministry of Agriculture and Rural Affairs as the training set, labeling samples as normal entities or risk entities based on historical actual operating conditions and risk records, and optimizing the regularization parameter C and the number of iterations max_iter of the logistic regression through 5-fold cross-validation and grid search method to determine the optimal parameters. S3: Identity verification and result output, including input of information to be verified, data association and retrieval, feature extraction and model calculation, information consistency verification, result judgment and output; S4: The multi-dimensional identity recognition and risk warning method also includes information security protection steps: in non-verification scenarios, sensitive information such as ID card numbers and social credit codes are displayed in anonymized form, and the database adopts role-based and permission-based hierarchical access control, so that only authorized personnel can view complete sensitive information.
[0008] Preferably, the data cleaning in S1 specifically involves deduplicating multiple registration records of the same business entity, filling in missing values in a differentiated manner, and filling missing values for annual net profit with the average net profit of family farms of the same main business type in the county over the past 3 years, and filling missing values for education level with 5-high school.
[0009] Preferably, the outlier handling method in S1 uses the Z-score method to correct fields with outliers, and the calculation formula is as follows: ; in, For field values, For the average value of the field, For the standard deviation of the field, Standard scores; For outliers in the fields of operating land area and annual total revenue that are more than three times the average of similar business entities in the county, replace them with boundary values that are ± three times the standard deviation of the average of the fields.
[0010] Preferably, the data standardization in S1 specifically involves using a normalization formula to map the numerical fields of operating land area, grain output, and annual total income to the interval of 0 and 1, and performing one-hot encoding on the classification fields of demonstration farm level and main business type. The normalization formula is as follows: ;in, The original value, and These are the minimum and maximum values of the field in the dataset, respectively.
[0011] Preferably, the data association method in S1 is to use the social credit code as the unique identifier of the business entity and the farm owner's ID number as the unique identifier of the legal representative, and connect the business data, identity data and credit data within the Ministry of Agriculture and Rural Affairs to form a business entity data archive containing full-dimensional information of the business entity.
[0012] Preferably, the identity correlation degree in S2 includes two sub-indicators: the consistency of the matching between the legal representative's ID number and the social credit code of the business entity in the registration system of the Ministry of Agriculture and Rural Affairs, and the overlap between the legal representative's mobile phone number and the contact number reserved by the business entity. Operational stability includes three sub-indicators: the change rate of operating land area in the past three years, the annual total revenue growth rate in the past two years, and the long-term proportion of transferred land. The level of business qualification is the demonstration farm level, with different scores corresponding to provincial, municipal, county and none levels in that order.
[0013] Preferably, the model construction in S2 uses a logistic regression algorithm, and the algorithm formula is: ;in, The score is the composite score in the range of 0 and 1. For identity relevance, For operational stability, For credit performance capability, To meet the level of business qualifications, - The weights are assigned as feature weights, and the weights are ranked as follows: identity association > operational stability > credit performance capability > operational qualification level.
[0014] Preferably, the result determination and output method in S3 is as follows: if the identity association degree meets the standard and the information consistency verification passes, the output "identity association passed" is output; otherwise, the output "identity association failed" is output. Based on the comprehensive score, four risk levels are divided, corresponding to A-level low risk, B-level medium risk, C-level high risk, and D-level extremely high risk, respectively.
[0015] Preferably, the information security protection steps are as follows: in non-verification scenarios, the middle 8 digits of the ID number are hidden, the last 4 digits of the social credit code are hidden, and the middle 4 digits of the contact number are hidden; The database is configured with three roles: agricultural and rural affairs bureau reviewer, system administrator, and data analyst. Reviewers can view complete sensitive information. Sensitive data transmission uses the SSL encryption protocol, and storage uses the AES-256 encryption algorithm.
[0016] A multidimensional identity recognition and risk warning system, comprising a data access module, a data preprocessing module, a feature engineering module, a logistic regression model module, an information verification module, and a risk warning output module; The data access module is used to connect to the family farm data interface of the Ministry of Agriculture and Rural Affairs database, collect raw data from all dimensions in real time and transmit it to the data preprocessing module. The data preprocessing module receives the raw data and then sequentially performs duplicate data removal, missing value filling, outlier correction, numerical normalization, classification coding, and dual primary key association operations, outputting the integrated business entity data archive to the feature engineering module. The feature engineering module is communicatively connected to the data preprocessing module, and is used to extract four core features and output standardized feature vectors, supporting automatic expansion of the independent encoding dimension of the classification field. The logistic regression model module is communicatively connected to the feature engineering module and is used to store the optimal model parameters and calculate the comprehensive score. The information verification module is independent of the rule engine of the model module. It only processes the consistency comparison of sensitive information and outputs the verification result of passing or failing to the risk warning output module. The risk warning output module is used to integrate the preliminary risk assessment results and the consistency verification results, and generate a final report containing the identity authentication conclusion, risk level, risk characteristics and reasons according to the preset template. The report is then fed back to the agricultural and rural affairs bureau user terminal through the system interface or API interface.
[0017] Compared with the prior art, the beneficial effects of the present invention are: 1. This law clarifies that by constructing a two-dimensional identity association system based on the legal representative's ID number and the business entity's social credit code, combined with cross-verification based on mobile phone number overlap, it breaks through the limitations of traditional single identity verification, accurately identifies situations where individuals and enterprises are separated, blocks the path for criminals to obtain financial funds from the source, significantly reduces the financial security risks caused by identity theft, and builds a solid security barrier for financial institutions' fund disbursement.
[0018] 2. This law clarifies that by integrating operational risk and credit risk for cross-analysis, and by using key indicators such as land area change rate and revenue growth rate to detect potential crises of operating entities in advance, compared with the traditional early warning model that only relies on overdue records, the risk identification cycle is advanced by 6-12 months, allowing financial institutions sufficient time to respond, significantly reducing idle and wasted funds, and improving the efficiency of capital allocation.
[0019] 3. This law clarifies that by replacing traditional manual verification with fully automated processing, the verification time for a single entity is reduced from several working days to tens of minutes, thus reducing waiting time and avoiding human error. In addition, through multiple security mechanisms such as anonymized display of sensitive information, hierarchical access control, and encrypted transmission and storage, sensitive data such as ID card numbers and social credit codes are strictly protected, thereby improving management efficiency and achieving reliable protection of sensitive information. Attached Figure Description
[0020] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0021] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system architecture diagram of the present invention; Figure 3 This is a flowchart of the process of the present invention. Detailed Implementation
[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0023] Please see Figures 1 to 3 The present invention provides a technical solution: Example 1, see Figure 1 and Figure 3 As shown, a multidimensional identity recognition and risk warning method is described. S1: Data collection and preprocessing. The data collection source is multidimensional data of family farms in the database of the Ministry of Agriculture and Rural Affairs. The multidimensional data of family farms includes the land area operated by the family farm, grain output, demonstration farm level, financial support funds, social credit code, registration status, legal representative's ID number, contact number, annual total income, and the area and term of land transfer. Preprocessing includes data cleaning, outlier handling, data standardization, and data association. The multidimensional data all come from the official database of the Ministry of Agriculture and Rural Affairs, specifically covering four categories: basic information of business entities, data on operating status, qualification and funding data, and supplementary information.
[0024] Basic information about the operating entity includes its social credit code, registration status, legal representative's ID number, contact number, and main business type. Operating status data includes the area of land operated in the past three years, grain output in the past three years, total annual income in the past two years, area and term of land transferred, and qualification and funding data including the demonstration farm level and amount of government support funds.
[0025] Supplementary information includes the farmer's education level and annual net profit.
[0026] S2: Feature engineering and model building. The core features extracted by feature engineering include identity correlation, business stability, credit performance capability, and business qualification level. The model construction adopts the logistic regression algorithm. The model training process includes selecting historical data of family farms in the database of the Ministry of Agriculture and Rural Affairs as the training set, labeling the samples as normal subjects or risk subjects according to the actual historical operation and risk records, and optimizing the regularization parameter C and the number of iterations max_iter of logistic regression through 5-fold cross-validation and grid search method to determine the optimal parameters. Historical data from 1,000 family farms in the database were selected as the training set and divided into a training subset and a test subset in a 7:3 ratio. Samples without identity theft, business crisis, or credit default were labeled as normal entities (label 1), while samples with any of these risks were labeled as risky entities (label 0).
[0027] During model training, the core feature weights that have been validated through training. =0.4、 =0.3、 =0.15、 =0.15, with the weights ranked as follows: identity relevance > operational stability > credit performance capability and operational qualification level. These four factors together constitute the core evaluation system of the model.
[0028] This method also includes a model iteration and optimization step. Every six months, the system automatically collects risk event data of family farms for the year, including cases of identity theft, business bankruptcy, and credit default, and adds them to the training set to retrain the logistic regression model and update the model parameters to improve prediction accuracy.
[0029] S3: Identity verification and result output, including input of information to be verified, data association and retrieval, feature extraction and model calculation, information consistency verification, and result determination and output; In the input field, users are asked to submit the legal representative's ID number and the business entity's social credit code.
[0030] Data association calls access multidimensional data from the Ministry of Agriculture and Rural Affairs through a secure interface, automatically linking it to the data archives of business entities.
[0031] Feature extraction and model calculation are based on extracting core features from the preprocessed data and substituting them into the trained logistic regression model to calculate the comprehensive score, which ranges from 0 to 1.
[0032] The information consistency verification compares the entered ID number and social credit code with the original data registered by the Ministry of Agriculture and Rural Affairs to confirm the consistency of the information.
[0033] The results are determined by combining the comprehensive score with the information consistency verification results, and the output includes the identity authentication result, risk level, and risk cause.
[0034] S4: The multi-dimensional identity recognition and risk warning method also includes information security protection steps. In non-verification scenarios, sensitive information such as ID card numbers and social credit codes are displayed in anonymized form. The database adopts role-based and permission-based hierarchical access control, and only authorized personnel can view complete sensitive information.
[0035] Example 2, see Figure 1 As shown, a multidimensional identity recognition and risk warning method is described. In S1, data cleaning specifically involves deduplicating multiple registration records of the same business entity, filling in missing values with differential values, and filling missing values for annual net profit with the average net profit of family farms of the same main business type in the county over the past three years, and filling missing values for education level with 5-high school.
[0036] Data deduplication is achieved through a dual primary key system using the social credit code and the legal representative's ID number, identifying and deleting duplicate records generated by multiple registrations of the same family farm due to information updates.
[0037] 5. High school is the mainstream level in the county. This value is used to fill in missing values for educational attainment to ensure that the data reflects the actual situation in the region. Outlier handling in S1 uses the Z-score method to correct fields with outliers. The calculation formula is as follows: ;in, For field values, For the average value of the field, For the standard deviation of the field, Standard scores; For outliers in the fields of operating land area and annual total revenue that are more than three times the average of similar business entities in the county, replace them with boundary values that are ± three times the standard deviation of the average of the fields.
[0038] Specifically, if the absolute value of the Z-score is greater than 3, the field is considered an outlier.
[0039] If the average land area operated by similar family farms is 200 mu (approximately 13.3 hectares) and the standard deviation is 150 mu (approximately 10.3 hectares), and a certain farm declares an area of 1000 mu (approximately 67 hectares), then the Z-score can be calculated as follows: The absolute value is greater than 3 and far exceeds the mean by 3 times. Replace it with the boundary value of 200 + 3 × 150 = 650 mu.
[0040] In S1, data standardization specifically involves using a normalization formula to map the numerical fields of operating land area, grain output, and annual total income to the 0,1 interval, and performing one-hot encoding on the fields of demonstration farm level and main business type classification. The normalization formula is as follows: ;in, The original value, and These are the minimum and maximum values of the field in the dataset, respectively.
[0041] Through formula Substituting the original values of numerical fields such as operating land area, grain output, and total annual income into this formula can uniformly map field values of different dimensions to the 0,1 range, eliminating the interference of dimensional differences on the subsequent training of the multidimensional identity recognition model. For example, assuming that the minimum value of operating land area in a certain county is 50 mu and the maximum value is 1000 mu, and the original area of a certain farm is 200 mu, then its normalized value is (200-50) / (1000-50)=150 / 950≈0.158, and the value can be directly used as the model input feature to participate in subsequent feature fusion and risk assessment calculations.
[0042] The data association method in S1 is to use the social credit code as the unique identifier of the business entity and the farm owner's ID number as the unique identifier of the legal representative, and to connect the business data, identity data and credit data within the Ministry of Agriculture and Rural Affairs to form a business entity data archive containing full-dimensional information about the business entity.
[0043] The social credit code corresponds to the business entity, and the farm owner's ID number corresponds to the legal representative. The data from multiple systems, including the Ministry of Agriculture and Rural Affairs' internal registration system and fund management system, are connected through a dual primary key.
[0044] For example, by associating social credit code 92130123MA0Gxxxx and farm owner ID number 130xxxxxxxxxxxxxxx, it is possible to integrate the family farm's land management data, fiscal fund usage data, and credit performance records to form a complete data archive of the business entity.
[0045] The identity correlation in S2 includes two sub-indicators: the consistency between the legal representative's ID number and the business entity's social credit code in the Ministry of Agriculture and Rural Affairs' registration system, and the overlap between the legal representative's mobile phone number and the business entity's reserved contact number. Operational stability includes three sub-indicators: the change rate of operating land area in the past three years, the annual total revenue growth rate in the past two years, and the long-term proportion of transferred land. The formula for the rate of change in the area of land under management over the past three years is: ; The formula for the annual total revenue growth rate over the past two years is: ; The formula for the long-term proportion of transferred land is: area with a transfer period of more than 5 years / total transferred area.
[0046] The level of business qualification is the demonstration farm level, with different scores corresponding to provincial, municipal, county and none levels in that order.
[0047] Among them, the demonstration farms are categorized into three levels: 0 (none), 1 (county level), 2 (city level), and 3 (provincial level). The higher the level, the better the qualification.
[0048] The model in S2 uses the logistic regression algorithm, and the algorithm formula is as follows: ; in, The score is the composite score in the range of 0 and 1. For identity relevance, For operational stability, For credit performance capability, To meet the level of business qualifications, - The feature weights are ranked as follows: identity relevance > operational stability > credit performance capability > operational qualification level. In S3, the result determination and output method is as follows: if the identity relevance meets the standard and the information consistency verification passes, the result is output as "identity relevance passed"; otherwise, the result is output as "identity relevance failed". Based on the comprehensive score, four risk levels are assigned: A (low risk), B (medium risk), C (high risk), and D (extremely high risk).
[0049] Among them, Grade A has stable operations and good credit; Grade B has minor problems with operations or credit; Grade C has contracted operations or defaulted on credit; and Grade D has severely deteriorated operations or identity theft.
[0050] The specific risk level classification criteria are as follows: Level A: Low risk (S≥0.8), Level B: Medium risk (0.6≤S<0.8), Level C: High risk (0.4≤S<0.6), and Level D: Extremely high risk (S<0.4).
[0051] The information security protection steps are as follows: in non-verification scenarios, the middle 8 digits of the ID number are hidden, the last 4 digits of the social credit code are hidden, and the middle 4 digits of the contact number are hidden; The database is configured with three roles: agricultural and rural affairs bureau reviewer, system administrator, and data analyst. Reviewers can view complete sensitive information. The database also includes regular users. System administrators can view all information, while auditors need to access complete sensitive information via username, password, and SMS verification. Data analysts use the data sources for statistical reports and risk warning trends, and these data sources must be automatically anonymized by the system, retaining only feature indicators related to multidimensional identity recognition and risk assessment, without including specific sensitive fields such as ID card numbers and contact numbers.
[0052] Ordinary users can only access the business data they submit and the de-identified public risk warning information, and cannot view any data or complete sensitive content of other users.
[0053] Sensitive data transmission uses the SSL encryption protocol, and storage uses the AES-256 encryption algorithm.
[0054] Example 3, see Figure 2 As shown, a multidimensional identity recognition and risk warning system includes a data access module, a data preprocessing module, a feature engineering module, a logistic regression model module, an information verification module, and a risk warning output module. The data access module is used to connect to the family farm data interface of the Ministry of Agriculture and Rural Affairs database, collect raw data from all dimensions in real time and transmit it to the data preprocessing module; The data preprocessing module receives the raw data and sequentially performs deduplication, missing value filling, outlier correction, numerical normalization, classification coding, and dual primary key association operations, outputting the integrated business entity data archive to the feature engineering module. The feature engineering module communicates with the data preprocessing module to extract four core features and output standardized feature vectors, supporting independent encoding of classification fields with automatic dimension expansion. The automatic feature update function of the feature engineering module is as follows: when the Ministry of Agriculture and Rural Affairs database adds a new classification field type, the system automatically identifies it and expands the dimension of its one-hot encoding to ensure the adaptability of feature extraction. The logistic regression model module communicates with the feature engineering module to store the optimal model parameters and calculate the comprehensive score; the information verification module is independent of the rule engine of the model module, only handles the consistency comparison of sensitive information, and outputs the verification result of pass or fail to the risk warning output module. The risk warning output module is used to integrate the preliminary risk assessment results and the consistency verification results, and generate a final report containing the identity authentication conclusion, risk level, risk characteristics and causes according to the preset template. The report is fed back to the agricultural and rural affairs bureau user terminal through three methods: system interface, API interface synchronization or PDF file export.
[0055] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A multi-dimensional identity recognition and risk warning method, characterized in that: The steps of the multi-dimensional identity recognition and risk warning method are as follows: S1: Data collection and preprocessing. The data collection source is multidimensional data of family farms in the database of the Ministry of Agriculture and Rural Affairs. The multidimensional data of family farms includes the land area operated by the family farm, grain output, demonstration farm level, financial support funds, social credit code, registration status, legal representative's ID number, contact number, annual total income, and the area and term of land transfer. The preprocessing includes data cleaning, outlier handling, data standardization, and data association. S2: Feature engineering and model building. The core features extracted by feature engineering include identity correlation, business stability, credit performance capability, and business qualification level. The model is constructed using a logistic regression algorithm. The model training process includes selecting historical data of family farms in the database of the Ministry of Agriculture and Rural Affairs as the training set, labeling samples as normal entities or risk entities based on historical actual operating conditions and risk records, and optimizing the regularization parameter C and the number of iterations max_iter of the logistic regression through 5-fold cross-validation and grid search method to determine the optimal parameters. S3: Identity verification and result output, including input of information to be verified, data association and retrieval, feature extraction and model calculation, information consistency verification, result judgment and output; S4: The multi-dimensional identity recognition and risk warning method also includes information security protection steps: in non-verification scenarios, sensitive information such as ID card numbers and social credit codes are displayed in anonymized form, and the database adopts role-based and permission-based hierarchical access control, so that only authorized personnel can view complete sensitive information.
2. The multi-dimensional identity recognition and risk warning method according to claim 1, characterized in that: The data cleaning in S1 specifically involves deduplicating multiple registration records of the same business entity, filling in missing values using differentiated methods, and filling missing values for annual net profit with the average net profit of family farms of the same main business type in the county over the past three years, and filling missing values for education level with 5-high school.
3. The multi-dimensional identity recognition and risk warning method according to claim 1, characterized in that: The outlier handling method in S1 uses the Z-score method to correct fields with outliers. The calculation formula is as follows: ； in, For field values, For the average value of the field, For the standard deviation of the field, The standard score is used; for outliers in the fields of operating land area and annual total income that are more than 3 times the average of similar business entities in the county, they are replaced with the boundary value of the mean of the field ± 3 times the standard deviation.
4. The multi-dimensional identity recognition and risk warning method according to claim 1, characterized in that: The data standardization in S1 specifically involves using a normalization formula to map the numerical fields of operating land area, grain output and annual total income to the 0,1 interval, and performing one-hot encoding on the demonstration farm level and main business type classification fields. The normalization formula is as follows: ； in, The original value, and These are the minimum and maximum values of the field in the dataset, respectively.
5. The multi-dimensional identity recognition and risk warning method according to claim 1, characterized in that: The data association method in S1 is to use the social credit code as the unique identifier of the business entity and the farm owner's ID number as the unique identifier of the legal representative, and to connect the business data, identity data and credit data within the Ministry of Agriculture and Rural Affairs to form a business entity data archive containing full-dimensional information of the business entity.
6. The multi-dimensional identity recognition and risk warning method according to claim 1, characterized in that: The identity correlation in S2 includes two sub-indicators: the consistency between the legal representative's ID number and the business entity's social credit code in the Ministry of Agriculture and Rural Affairs' registration system, and the overlap between the legal representative's mobile phone number and the business entity's reserved contact number. Operational stability includes three sub-indicators: the change rate of operating land area in the past three years, the annual total revenue growth rate in the past two years, and the long-term proportion of transferred land. The level of business qualification is the demonstration farm level, with different scores corresponding to provincial, municipal, county and none levels in that order.
7. The multi-dimensional identity recognition and risk warning method according to claim 1, characterized in that: The model in S2 uses a logistic regression algorithm, and the algorithm formula is as follows: ； in, The score is the composite score in the range of 0 and 1. For identity relevance, For operational stability, For credit performance capability, To meet the level of business qualifications, - The weights are assigned as feature weights, and the weights are ranked as follows: identity association > operational stability > credit performance capability > operational qualification level.
8. The multi-dimensional identity recognition and risk warning method according to claim 1, characterized in that: The result determination and output method in S3 is as follows: if the identity association degree meets the standard and the information consistency verification passes, the output "identity association passed" is output; otherwise, the output "identity association failed" is output. Based on the comprehensive score, four risk levels are divided, corresponding to A-level low risk, B-level medium risk, C-level high risk, and D-level extremely high risk, respectively.
9. The multi-dimensional identity recognition and risk warning method according to claim 1, characterized in that: The information security protection steps are as follows: in non-verification scenarios, the middle 8 digits of the ID number are hidden, the last 4 digits of the social credit code are hidden, and the middle 4 digits of the contact number are hidden. The database is configured with three roles: agricultural and rural affairs bureau reviewer, system administrator, and data analyst. Reviewers can view complete sensitive information. Sensitive data transmission uses the SSL encryption protocol, and storage uses the AES-256 encryption algorithm.
10. A multi-dimensional identity recognition and risk warning system, comprising a multi-dimensional identity recognition and risk warning method according to any one of claims 1-9, characterized in that: The multidimensional identity recognition and risk warning system includes a data access module, a data preprocessing module, a feature engineering module, a logistic regression model module, an information verification module, and a risk warning output module. The data access module is used to connect to the family farm data interface of the Ministry of Agriculture and Rural Affairs database, collect raw data from all dimensions in real time and transmit it to the data preprocessing module. The data preprocessing module receives the raw data and then sequentially performs duplicate data removal, missing value filling, outlier correction, numerical normalization, classification coding, and dual primary key association operations, outputting the integrated business entity data archive to the feature engineering module. The feature engineering module is communicatively connected to the data preprocessing module, and is used to extract four core features and output standardized feature vectors, supporting automatic expansion of the independent encoding dimension of the classification field. The logistic regression model module is communicatively connected to the feature engineering module and is used to store the optimal model parameters and calculate the comprehensive score. The information verification module is used to store the trained logistic regression model parameters, receive the feature vector, calculate the comprehensive score and preliminary risk label, and output the preliminary risk assessment result to the risk warning output module. The risk warning output module is used to integrate the preliminary risk assessment results and the consistency verification results, and generate a final report containing the identity authentication conclusion, risk level, risk characteristics and reasons according to the preset template. The report is then fed back to the agricultural and rural affairs bureau user terminal through the system interface or API interface.