Carbon trading market risk early warning system and method based on big data

Through the big data analysis system, combined with random forests and ResLSTM models, evaluation and early warning are solved, and the problem of insufficient future prediction capabilities of the carbon trading market in the existing technology is achieved, and effective prediction and early warning of carbon trading market risks is achieved.

CN116777213BActive Publication Date: 2024-11-08GUANGZHOU YICHUANG BRAND MANAGEMENT CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310741848.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-06-21
Publication Date
2024-11-08
Estimated Expiration
2043-06-21

AI Technical Summary

Technical Problem

The existing carbon trading model lacks the ability to predict the future of the carbon trading market and cannot effectively provide risk warnings.

Method used

Adopt a carbon trading market risk warning system based on big data, including data collection, data preprocessing, risk assessment and early warning modules, and use the random forest model and the long and short-term memory neural network model that introduces residual connections to conduct risk assessment and based on The evaluation results are issued with early warning.

Benefits of technology

It realizes prediction and early warning of risks in the carbon trading market, improves the accuracy and stability of risk prediction, and can more comprehensively capture market characteristics and trends, and cope with uncertainty and data diversity.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116777213B_ABST
    Figure CN116777213B_ABST
Patent Text Reader

Abstract

The present invention discloses a carbon trading market risk early warning system and method based on big data, the system includes a data acquisition module, a data preprocessing module, a risk assessment module, a model update module and an early warning module; the data acquisition module is connected to the data preprocessing module, the data preprocessing module is connected to the risk assessment module, the risk assessment module is connected to the early warning module, and the model update module is connected to the risk assessment module. The advantage of the present invention is that the risk assessment module includes a risk assessment model, the risk assessment model adopts a combination of a random forest model and a long short-term memory neural network model with residual connection, fully utilizing the advantages of the two to obtain a more comprehensive risk prediction capability.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of big data analysis, and in particular to a carbon trading market risk early warning system and method based on big data. Background Art

[0002] With the rise of the carbon trading market, more and more companies and individuals are participating in carbon trading, and the carbon trading market has become more complex. Therefore, it is necessary to conduct necessary risk warnings for the carbon trading market.

[0003] The current carbon trading model mainly uses carbon emission reduction, emission reduction benefits, clean energy utilization rate and other indicators to measure results. The existing carbon trading model lacks predictions for the future of the carbon trading market.

[0004] In summary, there is an urgent need for a carbon trading market risk early warning system and method based on big data to solve the problems existing in the existing technology. Summary of the invention

[0005] The purpose of the present invention is to provide a carbon trading market risk early warning system and method based on big data. The specific technical solution is as follows:

[0006] A carbon trading market risk early warning system based on big data, comprising a data acquisition module, a data preprocessing module, a risk assessment module, a model updating module and an early warning module; the data acquisition module is connected to the data preprocessing module, the data preprocessing module is connected to the risk assessment module, the risk assessment module is connected to the early warning module, and the model updating module is connected to the risk assessment module;

[0007] The data collection module is used to collect first transaction data;

[0008] The data preprocessing module is used to perform data cleaning on the first transaction data to obtain second transaction data;

[0009] The risk assessment module includes a risk assessment model, and the risk assessment model is used to perform risk assessment on the second transaction data to obtain a risk assessment result;

[0010] The model updating module is used to update and optimize the risk assessment model;

[0011] The warning module is provided with an adjustable risk threshold. When the risk assessment result exceeds the risk threshold, a warning is issued to the user, otherwise no warning is issued.

[0012] Preferably, in the data collection module, the first transaction data includes transaction data in the carbon trading market.

[0013] Preferably, in the data preprocessing module, the data cleaning includes processing missing values, processing abnormal values ​​and processing duplicate values.

[0014] Preferably, in the risk assessment module, the risk assessment model includes a random forest model and a long short-term memory neural network model with residual connection, wherein the random forest model predicts the second transaction data and obtains a first prediction result, the long short-term memory neural network model predicts the second transaction data and obtains a second prediction result, and the first prediction result and the second prediction result are weighted averaged to obtain a risk assessment result.

[0015] Preferably, the process of predicting the second transaction data by the random forest model is as follows:

[0016] Step A1: constructing a decision tree model, specifically, dividing the second transaction data into a training set and a test set, randomly extracting a certain number of samples from the training set with replacement to form multiple random subsets, randomly selecting features of each random subset to obtain multiple feature subsets, and constructing a decision tree model for the feature subsets using a classification and regression tree algorithm, wherein each feature subset corresponds to an independent decision tree model;

[0017] Step A2: Construct a random forest model. Specifically, multiple decision tree models are combined to obtain a random forest model. Samples in the test set are input into each decision tree model to obtain the prediction results of the decision tree model. The average of the prediction results of all decision tree models is the first prediction result.

[0018] Preferably, in step A2, a process of optimizing the random forest model is also included, specifically: calculating the root mean square error as an optimization index, adjusting the depth and the minimum number of samples of the decision tree model based on the optimization index, or adjusting the number of decision tree models and the size of the feature subset in the random forest model; the root mean square error expression is as follows:

[0019]

[0020] Among them, RMSE represents the root mean square error, N represents the number of samples, and y pred Represents the predicted value of the sample, y true Represents the true value of the sample.

[0021] Preferably, the process of predicting the second transaction data by the long short-term memory neural network model is as follows:

[0022] Step B1: constructing a long short-term memory neural network model, specifically, dividing the second transaction data into a training set, a validation set, and a test set, defining an input layer, a hidden layer, input layer nodes, hidden layer nodes, and an activation function of the long short-term memory neural network model, and adding residual connections between hidden layers; using mean square error as a loss function, and using a back propagation algorithm and an ADam optimization algorithm to train the long short-term memory neural network model to obtain a training model;

[0023] Step B2: Model validation, specifically, input the validation set into the training model to obtain the predicted value, and calculate the evaluation index based on the predicted value and the true label;

[0024] Step B3: Model tuning, specifically, tuning the training model according to the evaluation index, wherein the tuning includes adjusting the input layer, hidden layer, input layer nodes and hidden layer nodes of the training model, and selecting the best combination of the input layer, hidden layer, input layer nodes and hidden layer nodes by cross-validation to obtain a prediction model;

[0025] Step B4: Data prediction, specifically, inputting the test set into the prediction model, and calculating the second prediction result through the forward propagation process of the prediction model.

[0026] Preferably, in the early warning module, the early warning module continuously monitors the user's feedback, and if the user does not take any measures after receiving the early warning, the early warning is issued to the user again.

[0027] Preferably, the carbon trading market risk early warning system also includes a data visualization module for displaying risk assessment results to users in real time.

[0028] In addition, the present invention also discloses a carbon trading market risk early warning method based on big data, the method uses the carbon trading market risk early warning system as described above to implement the carbon trading market risk early warning method, the steps of the method are as follows:

[0029] Step S1: data collection, specifically, the data collection module collects first transaction data and transmits the first transaction data to the data preprocessing module;

[0030] Step S2: data cleaning, specifically, processing missing values, abnormal values ​​and duplicate values ​​in the first transaction data, then performing feature selection and normalization processing on the first transaction data to obtain second transaction data, and transmitting the second transaction data to the risk assessment module;

[0031] Step S3: risk assessment, specifically, inputting the second transaction data into a risk assessment model in a risk assessment module to obtain a risk assessment result, and transmitting the risk assessment result to the early warning module;

[0032] Step S4: Risk warning, specifically, the warning module detects whether the risk assessment result exceeds the preset risk threshold. When the risk assessment result exceeds the risk threshold, a warning is issued to the user, otherwise no warning is issued.

[0033] The application of the technical solution of the present invention has the following beneficial effects:

[0034] The present invention discloses a carbon trading market risk early warning system and method based on big data. The system includes a data acquisition module, a data preprocessing module, a risk assessment module, a model updating module and an early warning module. The data acquisition module acquires first transaction data in the carbon trading market, the data preprocessing module preprocesses the first transaction data to obtain second transaction data, the risk assessment module obtains a risk assessment result based on the second transaction data, and realizes the prediction of carbon trading market risks. The early warning module can issue an early warning to a user according to the risk assessment result.

[0035] The data preprocessing module in the present invention extracts features from the transaction data, and evaluates and screens the features using an information gain method, selecting features with higher information gain as the most important features.

[0036] The risk assessment module in the present invention includes a risk assessment model, which adopts a combination of a random forest model and a long short-term memory neural network model (ResLSTM model) with residual connections, making full use of the advantages of the two to obtain a more comprehensive risk prediction capability. The random forest model is good at processing structured data and feature importance ranking, and can be used for feature selection and prediction model construction; the random forest model predicts by integrating multiple decision trees, has a certain robustness to noise and outliers, and can handle complex relationships between multiple features, including nonlinear relationships and interaction effects, which solves the problem that risk estimation in the carbon trading market is affected by multiple factors. On the other hand, the present invention also introduces a residual connection LSTM model (ResLSTM model) to process time series data and capture long-term dependencies in the sequence. In the carbon trading market, time series data is of great significance, and risks and trends in the market often have temporal correlations. By introducing residual connections, the present invention can effectively alleviate the problem of gradient vanishing, enable the network to better capture long-term dependencies, and enhance the expressive power of the model, improve robustness and generalization capabilities, and promote the convergence and training efficiency of the model. By integrating the random forest model and the ResLSTM model, the present invention can more comprehensively capture the characteristics and trends of the carbon trading market, better cope with uncertainty and data diversity, and improve the accuracy and stability of risk prediction.

[0037] In addition to the above-described purposes, features and advantages, the present invention has other purposes, features and advantages. The present invention will be further described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings constituting a part of this application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0039] Figure 1 It is a system block diagram of the carbon trading market risk early warning system in the preferred embodiment of the present invention;

[0040] Figure 2 It is a unit structure flow chart of the LSTM model in the preferred embodiment of the present invention;

[0041] Figure 3 It is a flow chart of the steps of the carbon trading market risk early warning method in the preferred embodiment of the present invention. DETAILED DESCRIPTION

[0042] The embodiments of the present invention are described in detail below with reference to the accompanying drawings, but the present invention can be implemented in many different ways as defined and covered by the claims.

[0043] Example:

[0044] See also Figure 1 , a carbon trading market risk early warning system based on big data, comprising a data acquisition module, a data preprocessing module, a risk assessment module, a model updating module, an early warning module and a data visualization module; the data acquisition module is connected to the data preprocessing module, the data preprocessing module is connected to the risk assessment module, the risk assessment module is connected to the early warning module, and the model updating module is connected to the risk assessment module;

[0045] The data collection module is used to collect first transaction data; the first transaction data includes transaction data in the carbon trading market. In addition, the transaction data in this implementation may be carbon emission data, carbon trading price data or industry data.

[0046] The data preprocessing module is used to clean the first transaction data to obtain the second transaction data; specifically, the data cleaning includes processing missing values, processing abnormal values ​​and processing duplicate values; the data cleaning in this embodiment can check the integrity and accuracy of the first transaction data and exclude data that does not meet the requirements. In addition, the data preprocessing module can also unify the format of the data, which is conducive to subsequent risk assessment. After data cleaning, the data preprocessing module extracts characteristic variables related to the goal of carbon trading market risk prediction, and uses the information gain method to evaluate and screen the characteristic variables.

[0047] The risk assessment module includes a risk assessment model, which is used to perform risk assessment on the second transaction data to obtain a risk assessment result; the risk assessment model includes a random forest model and a long short-term memory neural network model that introduces residual connections, wherein the random forest model predicts the second transaction data and obtains a first prediction result, the long short-term memory neural network model predicts the second transaction data and obtains a second prediction result, and the first prediction result and the second prediction result are weighted averaged to obtain a risk assessment result.

[0048] The model update module is used to update and optimize the risk assessment model; this embodiment updates and optimizes the risk assessment model through the model update module to adapt to new data and business needs, improve the prediction performance and generalization ability of the model, solve the problem of the original model becoming invalid over time, and ensure the availability and accuracy of the model.

[0049] The warning module is provided with an adjustable risk threshold. When the risk assessment result exceeds the risk threshold, a warning is issued to the user, otherwise no warning is issued.

[0050] The data visualization module is used to display the risk assessment results to the user in real time.

[0051] Furthermore, the process of predicting the second transaction data by the random forest model in this embodiment is as follows:

[0052] Step A1: constructing a decision tree model, specifically, dividing the second transaction data into a training set and a test set, randomly extracting a certain number of samples from the training set with replacement to form multiple random subsets, randomly selecting features of each random subset to obtain multiple feature subsets, and constructing a decision tree model for the feature subsets using a classification and regression tree algorithm, wherein each feature subset corresponds to an independent decision tree model;

[0053] Step A2: Construct a random forest model. Specifically, multiple decision tree models are combined to obtain a random forest model. Samples in the test set are input into each decision tree model to obtain the prediction results of the decision tree model. The average of the prediction results of all decision tree models is the first prediction result.

[0054] Furthermore, the specific steps of building a decision tree model are as follows:

[0055] Step 1: Obtain a data set based on the second transaction data, and select an initial node based on the characteristics and labels of the data set;

[0056] Step 2: For each node, select a feature for partitioning, traverse all features, and for each feature, traverse all possible values ​​of the feature to divide the data set into subsets;

[0057] Step 3: Calculate the impurity of each subset. The classification and regression tree algorithm (CART algorithm) uses square error as a measurement indicator. The square error is defined as follows:

[0058]

[0059] Among them, MSE represents square error, N is the number of samples, y i is the true value of sample i, is the predicted value of sample i. The smaller the square error, the more accurate the prediction result.

[0060] Step 4: Select the minimum square error and determine the best partitioning features and partitioning points;

[0061] Step 5: If the subset after division meets the stopping condition, the node is marked as a leaf node, indicating a regression result of the decision tree model; if the subset after division does not meet the stopping condition, the node is marked as an internal node, and steps 2 to 4 are recursively repeated for each subset until the subset after division meets the stopping condition; the stopping condition is the preset decision tree depth and sample quantity;

[0062] In the sixth step, after building a complete decision tree model, the decision tree model is optimized by pruning some nodes or merging some leaf nodes through pruning operations (dropout) to improve the complexity and generalization ability of the decision tree model.

[0063] Furthermore, in step A2, the process of optimizing the random forest model is also included, specifically: calculating the root mean square error as an optimization index, adjusting the depth and the minimum number of samples of the decision tree model based on the optimization index, or adjusting the number of decision tree models and the size of the feature subset in the random forest model; the root mean square error expression is as follows:

[0064]

[0065] Among them, RMSE represents the root mean square error, N represents the number of samples, and y pred Represents the predicted value of the sample, y true Represents the true value of the sample.

[0066] Furthermore, the process of predicting the second transaction data using the long short-term memory neural network (LSTM) model in this embodiment is as follows:

[0067] Step B1: constructing a long short-term memory neural network model, specifically, dividing the second transaction data into a training set, a validation set and a test set, defining the input layer, hidden layer, input layer nodes, hidden layer nodes and activation function of the long short-term memory neural network model, and adding residual connections between the hidden layers; using mean square error as the loss function, and using the back propagation algorithm and the ADam optimization algorithm to train the long short-term memory neural network model to obtain a training model.

[0068] The residual connection specifically adds the output of the previous layer and the output of the current layer element by element to obtain the final residual connection output. The specific calculation formula is:

[0069] y res =y prev +y curr ;

[0070] Among them, y res Represents the residual connection output, y prev represents the output of the previous layer, y curr Represents the output of the current layer.

[0071] Furthermore, the specific calculation formula of the loss function is:

[0072]

[0073] Among them, Loss represents the loss, N is the number of samples, and y true,i is the true value of the sample, y pred,i is the sample prediction value.

[0074] The network is trained using the ADam (Adaptive moment estimation) optimization algorithm, which is an adaptive learning rate optimization algorithm that combines the characteristics of the momentum method and the adaptive learning rate. It dynamically adjusts the learning rate of each parameter by calculating the first-order moment estimate and the second-order moment estimate of the gradient.

[0075] The specific algorithm formula of ADam (Adaptive moment estimation) optimization algorithm is as follows:

[0076]

[0077] m and v represent the first-order moment estimate and the second-order moment estimate, respectively.

[0078] β 1 and β 2 are the exponential decay rates of the first-order moment estimate and the second-order moment estimate, respectively.

[0079] η is the learning rate

[0080] Step B2: Model validation, specifically, input the validation set into the training model to obtain the predicted value, and calculate the evaluation index based on the predicted value and the true label;

[0081] Step B3: Model tuning, specifically, tuning the training model according to the evaluation index, wherein the tuning includes adjusting the input layer, hidden layer, input layer nodes and hidden layer nodes of the training model, and selecting the best combination of the input layer, hidden layer, input layer nodes and hidden layer nodes by cross-validation to obtain a prediction model;

[0082] Step B4: Data prediction, specifically, inputting the test set into the prediction model, and calculating the second prediction result through the forward propagation process of the prediction model.

[0083] Furthermore, if Figure 2 The unit structure flow chart of the LSTM model shown:

[0084] ① Forget gate: accept a long-term memory C t-1 (the output from the previous unit module) and decide whether to keep or forget C t-1 Input the long-term memory at time t-1 into C t-1 Multiply by the forgetting factor f t The calculation formula of the forgetting factor is:

[0085] f t =σ(W f ·[h t-1 ,x t ]+b f );

[0086] ② Input gate: determines how much input information is saved to the cell state C at the current time t t The calculation formula is:

[0087] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0088]

[0089] And the cell state equation at time t:

[0090]

[0091] ③Output gate: controls the unit state C t How many outputs are there to the LSTM current output value h t The calculation formula is:

[0092] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0093] h t =o t ⊙tanh(C t );

[0094] Among them, x t Represents the current input data, which is composed of the training set divided by the above cross-validation method. 0 and b 0 denote the weight matrix and bias term respectively; σ(·) denotes the sigmoid function, and tanh(·) denotes the hyperbolic tangent function.

[0095] Preferably, in the early warning module, the early warning module continuously monitors the user's feedback, and if the user does not take any measures after receiving the early warning, the early warning is issued to the user again.

[0096] In addition, if Figure 3 As shown, this embodiment also discloses a carbon trading market risk early warning method based on big data, and the method applies the carbon trading market risk early warning system as described above to implement the carbon trading market risk early warning method, and the steps of the method are as follows:

[0097] Step S1: data collection, specifically, the data collection module collects first transaction data and transmits the first transaction data to the data preprocessing module;

[0098] Step S2: data cleaning, specifically, processing missing values, abnormal values ​​and duplicate values ​​in the first transaction data, then performing feature selection and normalization processing on the first transaction data to obtain second transaction data, and transmitting the second transaction data to the risk assessment module;

[0099] Step S3: risk assessment, specifically, inputting the second transaction data into a risk assessment model in a risk assessment module to obtain a risk assessment result, and transmitting the risk assessment result to the early warning module;

[0100] Step S4: Risk warning, specifically, the warning module detects whether the risk assessment result exceeds the preset risk threshold. When the risk assessment result exceeds the risk threshold, a warning is issued to the user, otherwise no warning is issued.

[0101] Furthermore, the feature selection is to select the feature variables related to the risk according to the goal of carbon trading market risk prediction, and then use the information gain method to evaluate and screen the features to determine the features with the most information for predicting carbon trading market risks. The features are sorted according to the size of the information gain, and the features with higher information gain are selected as the most important features. The expression is as follows:

[0102] G(D,A)=H(D)-H(D|A)

[0103]

[0104]

[0105] Among them, D is the data set, the sample size is |D|, K is the number of categories, |C k |Indicates category C k According to the value of feature A, D is divided into n subsets D 1 , D 2 ,……,D n . |D i | is sample D i The number of samples; G(D,A) is the information gain, H(D) is the empirical entropy of data set D; H(D|A) is the empirical conditional entropy of feature A on data set D.

[0106] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. The carbon trading market risk early warning system based on big data is characterized by: It includes a data acquisition module, a data preprocessing module, a risk assessment module, a model updating module and an early warning module; the data acquisition module is connected to the data preprocessing module, the data preprocessing module is connected to the risk assessment module, the risk assessment module is connected to the early warning module, and the model updating module is connected to the risk assessment module; The data collection module is used to collect first transaction data; The data preprocessing module is used to perform data cleaning on the first transaction data to obtain second transaction data; The risk assessment module includes a risk assessment model, and the risk assessment model is used to perform risk assessment on the second transaction data to obtain a risk assessment result; The model updating module is used to update and optimize the risk assessment model; The warning module is provided with an adjustable risk threshold. When the risk assessment result exceeds the risk threshold, a warning is issued to the user, otherwise no warning is issued; In the risk assessment module, the risk assessment model includes a random forest model and a long short-term memory neural network model with a residual connection, wherein the random forest model predicts the second transaction data and obtains a first prediction result, the long short-term memory neural network model predicts the second transaction data and obtains a second prediction result, and the first prediction result and the second prediction result are weighted averaged to obtain a risk assessment result; The process of predicting the second transaction data by the random forest model is as follows: Step A1: constructing a decision tree model, specifically, dividing the second transaction data into a training set and a test set, randomly extracting a certain number of samples from the training set with replacement to form multiple random subsets, randomly selecting features of each random subset to obtain multiple feature subsets, and constructing a decision tree model for the feature subsets using a classification and regression tree algorithm, wherein each feature subset corresponds to an independent decision tree model; Step A2: construct a random forest model, specifically, combine the constructed multiple decision tree models to obtain a random forest model, input the samples in the test set into each decision tree model, obtain the prediction results of the decision tree model, and the average value of the prediction results of all decision tree models is the first prediction result; In step A2, the process of optimizing the random forest model is also included, specifically: calculating the root mean square error as an optimization index, adjusting the depth and minimum number of samples of the decision tree model based on the optimization index, or adjusting the number of decision tree models and the size of feature subsets in the random forest model; the root mean square error expression is as follows: Among them, RMSE represents the root mean square error, N represents the number of samples, and y pre d Represents the predicted value of the sample, y true Represents the true value of the sample; The process of predicting the second transaction data by the long short-term memory neural network model is as follows: Step B1: constructing a long short-term memory neural network model, specifically, dividing the second transaction data into a training set, a validation set, and a test set, defining an input layer, a hidden layer, input layer nodes, hidden layer nodes, and an activation function of the long short-term memory neural network model, and adding residual connections between hidden layers; using mean square error as a loss function, and using a back propagation algorithm and an ADam optimization algorithm to train the long short-term memory neural network model to obtain a training model; Step B2: Model validation, specifically, input the validation set into the training model to obtain the predicted value, and calculate the evaluation index based on the predicted value and the true label; Step B3: Model tuning, specifically, tuning the training model according to the evaluation index, wherein the tuning includes adjusting the input layer, hidden layer, input layer nodes and hidden layer nodes of the training model, and selecting the best combination of the input layer, hidden layer, input layer nodes and hidden layer nodes by cross-validation to obtain a prediction model; Step B4: Data prediction, specifically, inputting the test set into the prediction model, and calculating the second prediction result through the forward propagation process of the prediction model.

2. The carbon trading market risk early warning system according to claim 1 is characterized in that: In the data collection module, the first transaction data includes transaction data in the carbon trading market.

3. The carbon trading market risk early warning system according to claim 1 is characterized in that: In the data preprocessing module, the data cleaning includes processing missing values, processing abnormal values ​​and processing duplicate values.

4. The carbon trading market risk early warning system according to claim 1 is characterized in that: In the early warning module, the early warning module continuously monitors the user's feedback, and if the user does not take any measures after receiving the early warning, the early warning is issued to the user again.

5. The carbon trading market risk early warning system according to claim 1 is characterized in that: It also includes a data visualization module to present risk assessment results to users in real time.

6. A carbon trading market risk early warning method based on big data, characterized in that: The carbon trading market risk early warning system according to any one of claims 1 to 5 is used to implement a carbon trading market risk early warning method, the steps of the method are as follows: Step S1: data collection, specifically, the data collection module collects first transaction data and transmits the first transaction data to the data preprocessing module; Step S2: data cleaning, specifically, processing missing values, abnormal values ​​and duplicate values ​​in the first transaction data, then performing feature selection and normalization processing on the first transaction data to obtain second transaction data, and transmitting the second transaction data to the risk assessment module; Step S3: risk assessment, specifically, inputting the second transaction data into a risk assessment model in a risk assessment module to obtain a risk assessment result, and transmitting the risk assessment result to the early warning module; Step S4: Risk warning, specifically, the warning module detects whether the risk assessment result exceeds the preset risk threshold. When the risk assessment result exceeds the risk threshold, a warning is issued to the user, otherwise no warning is issued.