A method for risk control modeling of rural customer data based on federated learning

Abstract

A method for modeling risk control of rural customer data based on federated learning includes the following steps: step 1, determining the dimensions of data indicators involved in training based on the data sources of the parties involved in the federated learning; step 2, performing data preprocessing and agreeing on the format of the training data; step 3, the parties need to perform dimensionality reduction processing on the data; step 4, the participants negotiate a specific agreement on model training; step 5, the participants initialize the model; step 6, federated learning training; step 7, repeating the training process in step 6 until a predetermined number of training rounds is reached or the loss function has been stable for a long time; the present invention relates to the field of data risk control technology. In the data preprocessing process, the data is preprocessed in an agreed manner, the original structure of the data is changed and the features are retained to the greatest extent, and the original data is not used for training, so that the data party is more likely to accept it, thereby obtaining more data sources.

Description

Technical Field

[0001] The present invention relates to the field of data risk control technology, and specifically to a data risk control modeling method for agriculture, rural areas and farmers based on federated learning. Background Art

[0002] Federated learning is an advanced machine learning method that allows model training to be performed in a distributed data environment without centralizing the data in one place. This is important for protecting privacy, reducing data transmission costs, and complying with data regulations.

[0003] The "Three Rural Issues" (agriculture, rural areas, and farmers) sector involves a vast amount of farmer and agricultural data. This data is often distributed across different regions or managed by different institutions, such as agricultural cooperatives and agricultural enterprises. Furthermore, agricultural data may contain a significant amount of sensitive information, such as farmers' personal information, land area, and crop varieties, which requires proper protection.

[0004] Logistic regression is a statistical learning method used to solve classification problems. It can be used to predict the probability of an event. In the field of financial risk management, logistic regression can be applied to tasks such as predicting whether a customer will fulfill their obligations.

[0005] The above technology still has the following problems: 1. The financial data generated by rural customers is relatively small, and they are in a situation where information is relatively congested. The amount of data from rural customers of a single institution is difficult to support risk control modeling.

[0006] 2. Banks and other financial institutions find it difficult to use their existing data to assess the credit status of rural customers, resulting in difficulties for rural customers to obtain loans. Even if they pass the assessment, the amount of loans they can get is very low.

[0007] 3. Financial institutions have blacklist data of customers and a large amount of basic information data of rural customers. However, banks and other financial institutions have strict data management regulations, which makes it difficult for the two institutions to exchange data, resulting in the problem of data islands.

[0008] 4. Federated learning technology is still developing rapidly, but there is still the problem of raw data being cracked through reverse engineering. If parties use raw data for training, there is still a risk of data leakage. Summary of the Invention

[0009] In view of the shortcomings of the existing technology, the present invention is implemented through the following technical solutions: a method for risk control modeling of rural customer data based on federated learning, comprising the following steps:

[0010] Step 1: Determine the data indicator dimensions involved in training based on the data sources of all parties involved in federated learning.

[0011] Step 2: Data preprocessing. Agree on the format for training data, and each participant preprocesses their own data, including data cleaning, missing value processing, and feature selection. For discrete variables, the random forest algorithm and the K-nearest neighbor algorithm are alternately used for filling in the gaps, while for continuous variables, the mean is used for filling in the gaps, ensuring that the data of all parties are relatively consistent in terms of features.

[0012] In step three, all parties need to perform dimensionality reduction on the data. For data with small differences within the same column, low-variance filtering is used, and for regular data, reverse feature elimination is performed. The processed data is divided into two parts, 85% of which is used for training, and each party retains 15% for verification.

[0013] Step 4: The participants negotiate a specific agreement on model training. Logistic regression is selected as the training model, and hyperparameters such as the training cycle and learning rate are determined based on the amount of data.

[0014] Step 5: Participants initialize the model.

[0015] Step 6: Federated Learning Training. This uses a combination of synchronous and asynchronous federation to ensure that all participants have sufficient time to provide data for model training. Model training steps include: local model update, model parameter aggregation, model aggregation, and feedback update.

[0016] Step 7: Repeat the training process in step 6 until the predetermined number of training rounds is reached or the loss function has stabilized for a long time.

[0017] In step 8, evaluate and validate the model. Use the reserved test data or conduct validation in real-world scenarios to evaluate the model's performance. If the accuracy is above 96%, use the model. Otherwise, increase the training cycle in step 4 by 10% each time. If the accuracy is still below 96%, reduce the learning rate in step 4 by 10%. Repeat the training until the accuracy reaches 96%.

[0018] This invention provides a data risk control modeling method for agriculture, rural areas and farmers based on federated learning. It has the following beneficial effects:

[0019] 1. Effectively protect privacy. Farmers' personal information is often sensitive information. Federated learning can be used to train models without transmitting data to a central server, thereby protecting data privacy.

[0020] 2. Break down data barriers and solve the problem of data silos. The data distribution of different institutions may be different. Federated learning allows model updates to be performed locally, avoiding the problem of centralizing data in one place.

[0021] 3. Reduced communication overhead: In federated learning, only the updated information of model parameters will be transmitted to the central server, rather than the original data, thereby reducing communication costs.

[0022] 4. Model reusability: Models trained using large data sets have good anti-interference capabilities and can be used repeatedly over a long period of time.

[0023] 5. During data preprocessing, use agreed methods to preprocess the data, change the original structure of the data and retain the features to the greatest extent possible. Do not use the original data for training, making it easier for the data party to accept and thus obtain more data sources. Implementation Method

[0024] The following is a clear and complete description of the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of the present invention.

[0025] The present invention provides a technical solution: according to the data sources of the parties involved in federated learning, the dimensions of the data indicators involved in training are determined.

[0026] Data preprocessing is performed, and the format of training data is agreed upon. Each participant preprocesses their own data, including data cleaning, missing value processing, feature selection, etc. For discrete variables, the random forest algorithm and the K-nearest neighbor algorithm are used alternately to fill in the gaps, and for continuous variables, the mean is used to fill in the gaps, ensuring that the data of all parties are relatively consistent in terms of features.

[0027] Each party needs to perform dimensionality reduction on the data. For data with small differences within the same column, low-variance filtering is used, and for regular data, reverse feature elimination is performed. The processed data is divided into two parts, 85% of which is used for training, and each party retains 15% for verification.

[0028] The participants negotiated a specific agreement on model training. Logistic regression was selected as the training model, and hyperparameters such as the training cycle and learning rate were determined based on the amount of data.

[0029] Participants initialize the model. Effective privacy protection is achieved. Farmers' personal information is often sensitive information. Federated learning can be used to train the model without transmitting the data to the central server, thus protecting data privacy.

[0030] Federated learning training uses a combination of synchronous and asynchronous federation to ensure that all participants have sufficient time to provide data for model training. Model training steps include local model updates, model parameter aggregation, model aggregation, and feedback updates. This breaks down data silos and addresses data silos. Data distribution may vary across institutions. Federated learning allows for local model updates, avoiding the problem of centralized data.

[0031] Repeat the training process in step 6 until the predetermined number of training rounds is reached or the loss function has stabilized for a long time. Model reusability: Models trained using large datasets have good anti-interference capabilities and can be reused over a long period of time.

[0032] During the data preprocessing process, the data is preprocessed using an agreed method to change the original structure of the data and retain the features to the greatest extent possible. The original data is not used for training, making it easier for the data party to accept and thus obtaining more data sources.

[0033] Evaluate and validate the model using reserved test data or in real-world scenarios. If the accuracy is above 96%, use the model. Otherwise, increase the training cycle in step 4 by 10% each time. If the accuracy is still below 96%, reduce the learning rate in step 4 by 10%. Repeat training until the accuracy reaches 96%.

[0034] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a..." does not exclude the presence of additional identical elements in the process, method, article, or device comprising the element.

[0035] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for risk control modeling of rural customer data based on federated learning, characterized by: The following steps are included: Step 1: Determine the data indicator dimensions involved in training based on the data sources of all parties involved in federated learning. Step 2: Data preprocessing. The format of training data is agreed upon, and each participant preprocesses their own data, including data cleaning, missing value processing, and feature selection. Step 3: All parties need to perform dimensionality reduction on the data. For data with small differences in the same column, low variance filtering is used for processing, and reverse feature elimination is performed on regular data. Step 4: The participants negotiate the specific agreement on model training; Step 5: Participants initialize the model; Step 6: Federated learning training, using a combination of synchronous and asynchronous federation to ensure that all participants have enough time to provide data for model training. Step 7: Repeat the training process in step 6 until the predetermined number of training rounds is reached or the loss function has stabilized for a long time; Step 8: Perform model evaluation and validation, using reserved test data or validating in real scenarios to evaluate the performance of the model.

2. The method for risk control modeling of rural customer data based on federated learning according to claim 1 is characterized by: In the step 2, the random forest algorithm and the K-nearest neighbor algorithm are alternately used to fill in the discrete variables, and the mean is used to fill in the continuous variables, so as to ensure that the data of each party are relatively consistent in terms of characteristics.

3. The method for risk control modeling of rural customer data based on federated learning according to claim 1 is characterized by: In step three, the processed data is divided, 85% of the data is used for training, and each party retains 15% of the data for verification.

4. The method for risk control modeling of rural customer data based on federated learning according to claim 1 is characterized by: In step 4, logistic regression is selected as the training model, and training cycle parameters and learning rate parameters are formulated according to the amount of data.

5. The method for risk control modeling of rural customer data based on federated learning according to claim 1 is characterized by: The model training steps in step six include: local model update, model parameter aggregation, model aggregation, and feedback update.

6. The method for risk control modeling of rural customer data based on federated learning according to claim 1 is characterized by: If the accuracy rate in step 8 is higher than 96%, this model is used. Otherwise, the training cycle in step 4 needs to be increased by 10% each time. If the accuracy rate is still lower than 96%, the learning rate parameter in step 4 is reduced by / 10 each time, and the training is repeated until the accuracy rate reaches 96%.