Intelligent financial risk monitoring method and system based on fusion regression

Abstract

The present invention discloses an intelligent financial risk monitoring method and system based on fusion regression, comprising the following steps: a system terminal collects and monitors historical data of financial products; extracts key factors; establishes time series data of key factors; machine learning; establishes regression models of financial product returns on different key factor data; establishes time regression models of key factors; obtains prediction results of financial product return data from different key factors; sets weight inclination and performs weighted fusion on multiple prediction results; compares with risk thresholds and the system makes risk judgments; fuses multiple prediction results by different key factors, optimizes the prediction results, and realizes real-time monitoring of financial risks; provides a modal decomposition method to realize optimized prediction of each key factor; can capture accidental factors affecting the returns of financial products, and provides a method for modal decomposition prediction and comparison of financial product returns to improve prediction accuracy.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence technology, and specifically relates to an intelligent financial risk monitoring method and system based on fusion regression. Background Art

[0002] Financial risk refers to risks related to finance, such as financial market risk, financial product risk, financial institution risk, etc. With the development of social economy, financial products are diversified on the market. For example, banks will also launch a large number of financial products. While financial products bring returns to investors, they also have certain volatility risks.

[0003] There are many factors that affect the volatility of financial products, such as housing prices, exchange rates, inflation rates, etc., which can cause volatility risks in financial products. For investors, financial product risks can affect their normal business activities and even bankruptcy, and will also affect the credibility of providers. How to control the volatility of financial products and then predict the risks of financial products has always been a difficult problem.

[0004] With the rapid development of artificial intelligence, relevant monitoring through big data has been applied to various industries, and is currently also applied to the field of financial product risk monitoring. At present, financial risk monitoring through artificial intelligence generally analyzes the historical transaction data and related influencing factors of financial products, and then predicts the returns of financial products through big data learning and influencing factors.

[0005] However, the current financial product risk monitoring has the following problems: 1. The prediction accuracy of influencing factors on financial product returns is low; 2. The prediction accuracy of influencing factors themselves and their weights is low; 3. There is a lack of prediction and monitoring of incidental factors; 4. The prediction accuracy of financial product returns is low when relying solely on influencing factors. Summary of the invention

[0006] The present invention provides an intelligent financial risk monitoring method and system based on fusion regression, which achieves real-time monitoring of financial risks by fusing multiple prediction results of different key factors and optimizing the prediction results; provides a modal decomposition method to achieve optimized prediction of each key factor; can capture accidental factors that affect the returns of financial products, and provides a method for modal decomposition prediction and comparison of financial product returns to improve prediction accuracy, so as to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solution: an intelligent financial risk monitoring method based on fusion regression, comprising the following steps: S1: The system terminal collects and monitors the historical data of financial products; S2: Extract feature data, form a data set, and extract key factors; during this period, establish time series data of key factors respectively; S3: Perform machine learning on the key factor model; at the same time, perform machine learning on the time series data for establishing the key factors; S4: Use machine learning to establish regression models of financial product returns for different key factor data; by establishing time series data of key factor data, decompose them according to the characteristic factors corresponding to the key factors, find the time series data of its characteristic factors, and establish time regression models of key factors through machine learning; S5: Predict the regression models of different key factors, and obtain the prediction results of different key factors on the financial product income data based on the regression models of different key factors and financial product income; S6: After the prediction results are obtained, the next fusion regression program is entered to trigger the fusion regression operation and set the weight tilt, that is, the corresponding weight tilt of different key factors is obtained through other external factors, and multiple prediction results are weightedly fused; S7: Compare the fusion regression results with the set risk threshold, and the system makes risk judgment; S8: Then, according to the judgment result, dynamic indication is given on the system terminal display.

[0008] Preferably, after risk assessment is performed, an intelligent control program is also linked, which includes system terminal control and mobile terminal operation. The mobile terminal operation includes web page and app operations, and corresponding control operations are performed to provide operational instructions after multi-level assessment of financial risks.

[0009] Preferably, in step S2, the data set is divided into a data set and a validation set, and the data set is divided into a training set and a test set, the training set is divided into different parts, and the test set takes the remaining parts of the data set according to the situation of the training set, and regression models with different parameters are obtained from different parts of the training set through learning, and the regression models with different parameters are tested with the test set to obtain test results of different regression models, the average error is calculated, and the model with the smallest error is selected and then verified with the validation set to achieve re-optimization of the model parameters.

[0010] Preferably, the specific steps of establishing the key factor regression model for the data set in step S2 are: establishing the key factor time series data of the data set, performing modal decomposition on the key factor time series data, and finding the corresponding characteristic components of each key factor, that is, the constituent components of each key factor; performing time series data within the corresponding period, and then obtaining a curve of each component time and corresponding value, and then machine training and learning to obtain a prediction of each component, and the prediction results of all components are calculated according to the formula of key factors and constituent components, and then a regression model of each key factor with respect to time is obtained.

[0011] Preferably, the weight tilt is set in S5 to include a weight tilt factor for similar financial products, a data set is established through similar financial products, time series data of key factors are obtained, a time curve of key factors is obtained, and fluctuation points are found, and the incidental influencing factors that affect the key factors on the time line of the fluctuation point are selected through big data analysis; and further a preliminary weight division of the degree of influence of the key factors is performed, and weight values ​​are set according to the size of the influence and the positive and negative correlation, and a weight influence period is set according to the time span of the influence; after selecting the time interval for predicting the return of the financial product, the above factors within the time interval are matched, and then the weight values ​​within the time interval are matched, and the weight values ​​of the interval are added positively and negatively to obtain the influence weight values ​​of different key factors, that is, to obtain weight tilt factors of different key factors, and before the prediction results of different key factors are fused and regressed, the weights of different key factors are comprehensively weighted with weight tilt factors, and weight tilt is performed to predict the fusion regression results.

[0012] Preferably, weight tilting is set in S5, including obtaining time series data of financial product returns, and then performing modal decomposition on the component characteristic parts of the financial product returns; then performing time series data for each component characteristic factor within the corresponding period, and then obtaining a curve of each component time and corresponding value, and then machine training and learning to obtain a prediction of each component, all component prediction results are calculated according to the formula of financial product returns and constituent components, and then a regression model of financial product returns with respect to time is obtained, and the model is combined with the regression result prediction in S6 for weight tilting.

[0013] Preferably, different key factors are dynamically iterated and divided into periods, and the financial product returns are compared with the prediction results of the corresponding financial product return regression model with respect to time based on the prediction results of different periods, so as to find the key factor with the minimum error, perform the weight tilt of the current period, and then summarize the weight tilt of each period, and optimize the weight of the prediction result of the fusion regression result of the prediction results of different key factors.

[0014] Preferably, different key factors are dynamically iterated and divided into periods, and the financial product returns are compared with the prediction results of the corresponding financial product return regression model on time based on the prediction results of different periods, and the error values ​​between the different key factors and the financial product return regression model on time are found, and the error ratio is calculated. The weights of different key factors in the current period are tilted, and then the weight tilts of each period are summarized, and the prediction results of different key factors are fused with the regression result prediction to optimize the weights.

[0015] The present invention provides a system for an intelligent financial risk monitoring method applied to fusion regression, comprising a system terminal, wherein the system terminal comprises a data management module, an operation module, a monitoring module and a display module; the data management module comprises a database, a data acquisition module, a machine learning module, a data processing module and a data analysis module; the database is used to store historical data related to financial product returns; the data acquisition module is used to collect historical data related to financial product returns; the machine learning module is used for training and learning to establish a financial product return model; the data processing module is used for the calculation of each regression function; the data analysis module is used for data comparison and risk judgment; the operation module comprises a terminal control module and a mobile app / webpage module, which are used for data operation, investment adjustment, etc.; the monitoring module is used for real-time monitoring of risk analysis results; and the display module is used for displaying reminders of risk conditions.

[0016] Compared with the prior art, the present invention has the following beneficial effects: 1. Predict multiple results through different key factors, introduce fusion regression equation to fuse multiple results, optimize the prediction results, and then realize real-time monitoring of financial risks.

[0017] 2. Provide a modal decomposition method to find the components of each key factor, and then aggregate and calculate multiple components to achieve optimized prediction of each key factor, and combine the regression model of returns on key factors to obtain the return prediction of financial products for different key factors.

[0018] 3. It can capture the accidental factors that affect the returns of financial products and make corresponding weight adjustments to further improve the stability of the model and the prediction accuracy.

[0019] 4. Provide a method to compare the modal decomposition prediction of financial product returns with the prediction of financial product returns based on key factors, and continue to optimize the corresponding weight tilt to further improve the prediction accuracy.

[0020] 5. Provide a system that optimizes the forecast of financial product returns by integrating regression and can further optimize the forecast by performing corresponding weight tilting. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 It is a schematic diagram of the intelligent control process of the present invention; Figure 3 It is a schematic diagram of the regression model optimization process of the present invention; Figure 4 It is a schematic diagram of modal decomposition of key factors of the present invention; Figure 5This is a schematic diagram of the weight tilting process of similar financial products of the present invention; Figure 6 This is a schematic diagram of the modal decomposition of the financial product income of the present invention. Figure 1 ; Figure 7 This is a schematic diagram of the modal decomposition of the financial product income of the present invention. Figure 2 ; Figure 8 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe 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 creative work are within the scope of protection of the present invention.

[0023] See also Figure 1 The present invention provides an intelligent financial risk monitoring method based on fusion regression, comprising the following steps: S1: The system terminal collects and monitors the historical data of financial products; automatically captures and analyzes historical data such as price fluctuation information, transaction information, housing prices, exchange rates, inflation rates, etc. related to the returns of financial products through the collection port; such as identifying the public data source of the network link of financial data through the API interface and web crawler, such as the financial website that publishes the data, etc., sending a request to the target website through the HTTP protocol, parsing it using the parsing library, and saving the extracted data to the database; Data includes structured data and unstructured data. Unstructured data can be converted into structured data through conventional tools. This not only improves the efficiency of data collection, but also facilitates the dynamic and continuous collection of data. Converting unstructured data into structured data facilitates subsequent processing and analysis. S2: Extract feature data and key factors; clean the data, process outliers and missing values ​​in the data, and conduct preliminary evaluation of the data in a database in the analysis environment, preferably a SQL database, to check the structure, type, missing values, etc. of the data, and then process missing values ​​and outliers, convert data types and adjust formats to form a data set; According to the set procedure, the system divides the data into data set a and data set b, sets the financial product income as the dependent variable, finds out the key factors affecting the dependent variable, selects the main fixed independent variables, such as house prices, exchange rates, and inflation rates, finds out the financial product income data and house prices, exchange rates, and inflation rate data within a specific period of the past financial products; and establishes the time series data of house prices, exchange rates, and inflation rates during the period; S3: Machine learning; select the model type through the financial product income data within the selected period and the house price, exchange rate, and inflation rate data within the period, use the API provided by the framework to define the model's hierarchy and connection method, configure the model's conventional optimizer and loss function, and perform machine learning training features; at the same time, perform machine learning on the time series data of house prices, exchange rates, and inflation rates; S4: Use machine learning to establish a regression model of financial product returns on housing prices, exchange rates, and inflation rates; specifically, set housing prices, exchange rates, and inflation rates to different x, i.e., x1, x2, and x3, and use machine learning to find the correlation between financial product returns y and x1, x2, and x3, and then derive the regression function between y and x1, x2, and x3 to establish a regression model; for example, the model selects a univariate regression equation; y=β0+β1x1+ε; y=β0+β1x2+ε; y=β0+β1x3+ε; β0 represents the intercept term, which is the expected value of y in the absence of an independent variable; β1 represents the coefficient, and ε represents the error term; By establishing time series data of housing prices, exchange rates, and inflation rates, we decompose them according to the characteristic factors corresponding to the key factors. This step is to take housing prices as the dependent variable, find the time series data of its characteristic factors, and establish time regression models of housing prices, exchange rates, and inflation rates through machine learning. S5: Since the regression models of housing prices, exchange rates, and inflation rates can be predicted in a set time period in the future, short-term prediction is preferred. The shorter the time, the higher the prediction accuracy. In this way, housing prices, exchange rates, and inflation rates can be predicted within the prediction period, and regression models can be formed based on different key factors and financial product returns to obtain prediction results of different housing prices, exchange rates, and inflation rates on financial product return data; S6: After the prediction result, different x1, x2, and x3 are found through model decomposition, and then substituted into the univariate regression equation to obtain different y1, y2, and y3. After the result is set, enter the next fusion regression program to trigger the fusion regression operation; z=(y1×weight1+y2×weight2+y3×weight3); z represents the fusion regression prediction result, weight 1 represents the weight of the key factor x1, weight 2 represents the weight of the key factor x2, and weight 3 represents the weight of the key factor x3; here, the weight tilt is set, that is, the corresponding weight tilt of x1, x2, and x3 is obtained through other external factors; The present invention performs fusion regression by setting a weighted average operation, substitutes y1, y2, and y3 into the weighted average function, and performs weighted fusion on multiple y1, y2, and y3; S7: Compare the fusion regression results with the set risk threshold, and the system makes risk judgment; S8: Then, according to the judgment result, a dynamic indication is given on the system terminal display. If the fusion regression result, that is, the expected return of the financial product, is lower than the risk threshold, a red warning is issued; if the fusion regression result, that is, the expected return of the financial product, is higher than the risk threshold, a green prompt is issued.

[0024] See also Figure 2 As a further embodiment, after risk judgment is made, an intelligent control program is also linked, which includes system terminal control and mobile terminal operation. The mobile terminal operation includes the operation of web pages and apps, and performs corresponding control operations, such as investment adjustments, etc.; the operation through the mobile terminal must be reviewed by the system terminal, and the risk warning setting has a prompt for the user to perform system operations; and the operation controlled by the system terminal is an operation performed according to the risk status with the user's authorization in advance, and then the system terminal will display a reminder after the operation, while unauthorized users can operate the app mobile terminal, and then the system terminal will review and perform corresponding operations on financial products, so as to provide operation instructions after multi-level judgment of financial risks. The instructions here are guidelines for the actual operation of financial products, which are conducive to the user's corresponding subsequent operations based on the actual financial products.

[0025] See also Figure 3 As a further embodiment, the data set a in step S2 is divided into a data set c and a validation set, and the data set c is divided into a training set and a test set, the training set is divided into different parts, and the test set takes the remaining parts of the data set c according to the situation of the training set, and regression models with different parameters are obtained by learning different parts of the training set, and the regression models with different parameters are tested with the test set to obtain test results of different regression models, the average error is calculated, and the model with the smallest error is selected for verification with the validation set to achieve re-optimization of the model parameters.

[0026] See also Figure 4 As a further embodiment, the specific steps of establishing the key factor regression model for the data set b in step S2 are: establishing the key factor time series data of the data set b, performing modal decomposition on the key factor time series data, and finding the corresponding characteristic components of each key factor, that is, the components that constitute the housing price, exchange rate, and inflation rate respectively; for example, the components of the housing price include land price, building price, sales cost, and transaction tax; the components of the exchange rate include foreign currency amount and local currency amount; the components of the inflation rate include the current price level and the base period price level; performing time series data within the corresponding period, and then obtaining a curve of each component time and corresponding value, and then machine training and learning to obtain the prediction of each component, the prediction results of all components are calculated according to the formula of key factors and components, and then the regression model of each key factor with respect to time is obtained.

[0027] See also Figure 5 As a further embodiment, the weight tilt is set in S5, including the weight tilt factor of similar financial products. A data set d is established through similar financial products to obtain the time series data of the key factors, the time curve of the key factors, and the fluctuation point is found. The accidental influencing factors that affect the key factors on the time line of the fluctuation point are selected through big data analysis, such as epidemics, policies and other factors; and the initial weight division of the degree of influence of the key factors is further performed, and the weight value is set according to the influence size and the positive and negative correlation. The weight influence period is set according to the time span of the influence; after selecting the time interval for predicting the return of the financial product, the above factors in the time interval are matched, and whether there are epidemics, policies and other factors at present are matched through big data. If there are, the relevant network data for the current epidemic and policy size assessment is linked. If not, the relevant network data for the possibility and size assessment of the above accidental factors are linked, and then the weight value in the time interval is matched, and the weight value of the interval is added positively and negatively to obtain the influence weight value of different key factors, that is, the weight tilt factor of different key factors is obtained. Before the prediction results of different key factors are fused and regressed, the weights of different key factors are comprehensively weighted and tilted, and the fusion regression results are predicted.

[0028] See also Figure 6 As a further embodiment, weight tilting is set in S5, including obtaining time series data of financial product returns, and then performing modal decomposition on the component characteristic parts of the financial product returns. Assuming that the financial product returns include component characteristics such as market interest rates and interest rate differentials, the market interest rate plus the interest rate differential is required to obtain the financial product yield, and further obtain the returns; then the time series data of each component characteristic factor within the corresponding period is performed, and then a curve of each component time and corresponding value can be obtained, and then machine training and learning are performed to obtain a prediction of each component. All component prediction results are calculated according to the formula of financial product returns and constituent components, and then a regression model of financial product returns with respect to time is obtained, and the model is combined with the regression result prediction in S6 for weight tilting.

[0029] See also Figure 6 As a further embodiment, different key factors are dynamically iterated and phased. For example, to predict the returns in the next six months, the six months are divided into six months, and the returns of financial products are predicted by different key factors each month. The financial product returns are compared with the prediction results of the corresponding financial product return regression model with respect to time based on the prediction results of different periods, and the key factor with the minimum error is found. The weight tilt of the current period is performed, and the comparison of the next period is performed in the same way. The weight tilt of the next month is performed until the last month is calculated, and the weight tilt of the last month is performed. Then, the weight tilt of each period is summarized, and the weight of the prediction results of different key factors is optimized by integrating the regression results, so as to achieve high prediction accuracy.

[0030] See also Figure 7 As a further embodiment, different key factors are dynamically iterated and phased. For example, to predict the earnings in the next six months, the six months are divided into six months, and the earnings of financial products are predicted by different key factors each month. The earnings of financial products are compared with the prediction results of the corresponding regression model of the earnings of financial products on time based on the prediction results of different periods, and the error values ​​between the different key factors and the regression model prediction of the earnings of financial products on time are found, and the error ratio is calculated. The weights of different key factors in the current period are tilted. Similarly, the comparison of the next period is performed, and the weight tilt of the next month is performed until the calculation reaches the last month. The weight tilt of the last month is performed, and then the weight tilt of each period is summarized, and the prediction results of different key factors are fused. The regression result prediction is optimized to achieve higher prediction accuracy, but the amount of calculation is larger than that of the previous embodiment, which can be selected in combination with actual conditions.

[0031] See also Figure 8 The present invention provides an intelligent financial risk monitoring system based on fusion regression, including a system terminal, wherein the system terminal includes a data management module, an operation module, a monitoring module and a display module; the data management module includes a database, a data acquisition module, a machine learning module, a data processing module and a data analysis module; the database is used to store historical data related to financial product returns; the data acquisition module is used to collect historical data such as price fluctuation information, transaction information, housing prices, exchange rates, inflation rates, etc. related to financial product returns; the machine learning module is used for training and learning of financial product returns and key factor models, training and learning of key factor time series models, and training and learning of financial product return time series models; the data processing module is used for the calculation of various regression functions; the data analysis module is used for data comparison and risk judgment; the operation module includes a terminal control module and a mobile app / webpage module, which are used for data operation, investment adjustment, etc.; the monitoring module is used for real-time monitoring of risk analysis results; and the display module is used for displaying reminders of risk conditions.

[0032] Although 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 the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent financial risk monitoring method based on fusion regression, characterized in that: The steps include: S1: The system terminal collects and monitors the historical data of financial products; S2: Extract feature data, form a data set, and extract key factors; during this period, establish time series data of key factors respectively; S3: Perform machine learning on the key factor model; at the same time, perform machine learning on the time series data for establishing the key factors; S4: Use machine learning to establish regression models of financial product returns for different key factor data; by establishing time series data of key factor data, decompose them according to the characteristic factors corresponding to the key factors, find the time series data of its characteristic factors, and establish time regression models of key factors through machine learning; S5: Predict the regression models of different key factors, and obtain the prediction results of different key factors on the financial product income data based on the regression models of different key factors and financial product income; S6: After the prediction results are obtained, the next fusion regression program is entered to trigger the fusion regression operation and set the weight tilt, that is, the corresponding weight tilt of different key factors is obtained through other external factors, and multiple prediction results are weightedly fused; S7: Compare the fusion regression results with the set risk threshold, and the system makes risk judgment; S8: Then, according to the judgment result, dynamic indication is given on the system terminal display.

2. The intelligent financial risk monitoring method based on fusion regression according to claim 1 is characterized in that: After making a risk assessment, an intelligent control program is also linked. The intelligent control program includes system terminal control and mobile terminal operation. Mobile terminal operation includes web page and app operation, which performs corresponding control operations to provide operational instructions after multi-level assessment of financial risks.

3. The intelligent financial risk monitoring method based on fusion regression according to claim 1 is characterized in that: In step S2, the data set is divided into a data set and a validation set, and then into a training set and a test set. The training set is divided into different parts, and the test set takes the remaining parts of the data set according to the situation of the training set. Different parts of the training set are used to obtain regression models with different parameters through learning. The regression models with different parameters are tested with the test set to obtain test results of different regression models, and the average error is calculated. The model with the smallest error is selected and then verified with the validation set to achieve re-optimization of model parameters.

4. The intelligent financial risk monitoring method based on fusion regression according to claim 1 is characterized in that: The specific steps for establishing the key factor regression model for the data set in step S2 are: establish the key factor time series data of the data set, perform modal decomposition on the key factor time series data, find the corresponding characteristic components of each key factor, that is, the constituent components of each key factor; perform time series data within the corresponding period, and then obtain the curve of each component time and corresponding value, and then perform machine training and learning to obtain the prediction of each component. The prediction results of all components are calculated according to the formula of key factors and constituent components, and then the regression model of each key factor with respect to time is obtained.

5. The intelligent financial risk monitoring method based on fusion regression according to claim 1 is characterized in that: The weight tilting set in S5 includes the weight tilting factor of similar financial products. A data set is established through similar financial products to obtain the time series data of key factors, the time curve of key factors, and the fluctuation point. The accidental influencing factors that affect the key factor on the time line of the fluctuation point are selected through big data analysis; and further the initial weight division of the degree of influence of the key factor is carried out, and the weight value is set according to the influence size and the positive and negative correlation. The weight influence period is set according to the time span of the influence; after selecting the time interval for predicting the return of the financial product, the above factors within the time interval are matched, and then the weight value within the time interval is matched, and the weight value of the interval is added positively and negatively to obtain the influence weight value of different key factors, that is, the weight tilting factors of different key factors are obtained. Before the prediction results of different key factors are fused and regressed, the weights of different key factors are comprehensively integrated with the weight tilting factors, and weight tilt is performed to predict the fusion regression results.

6. The intelligent financial risk monitoring method based on fusion regression according to claim 1 is characterized in that: Weight tilting is set in S5, including obtaining the time series data of financial product returns, and then performing modal decomposition on the component characteristic parts of the financial product returns; then performing time series data on each component characteristic factor within the corresponding period, and then obtaining a curve between each component time and corresponding value, and then machine training and learning to obtain a prediction of each component. The prediction results of all components are calculated according to the formula of financial product returns and constituent components, and then a regression model of financial product returns with respect to time is obtained, and the model is combined with the regression result prediction in S6 for weight tilting.

7. The intelligent financial risk monitoring method based on fusion regression according to claim 6 is characterized in that: Different key factors are dynamically iterated and divided into periods. The financial product returns are compared with the prediction results of the corresponding financial product return regression model with respect to time based on the prediction results of different periods. The key factors with the minimum error are found, and the weight tilt of the current period is performed. Then the weight tilt of each period is summarized, and the prediction results of different key factors are integrated to optimize the weight of the regression result prediction.

8. The intelligent financial risk monitoring method based on fusion regression according to claim 6 is characterized in that: Different key factors are dynamically iterated and divided into periods. The financial product returns are compared with the prediction results of the corresponding financial product return regression model on time based on the prediction results of different periods. The error values ​​between different key factors and the financial product return regression model on time are found, and the error ratio is calculated. The weights of different key factors in the current period are tilted, and then the weight tilts of each period are summarized. The prediction results of different key factors are integrated and the regression results are predicted to optimize the weights.

9. A system for the intelligent financial risk monitoring method of fusion regression according to any one of claims 1 to 8, characterized in that: It includes a system terminal, which includes a data management module, an operation module, a monitoring module and a display module; the data management module includes a database, a data acquisition module, a machine learning module, a data processing module and a data analysis module; the database is used to store historical data related to financial product returns; the data acquisition module is used to collect historical data related to financial product returns; the machine learning module is used for training and learning to establish a financial product return model; the data processing module is used for the calculation of various regression functions; the data analysis module is used for data comparison and risk judgment; the operation module includes a terminal control module and a mobile app / webpage module, which are used for data operation, investment adjustment, etc.; the monitoring module is used for real-time monitoring of risk analysis results; and the display module is used for displaying reminders of risk conditions.

Images