Financial risk control system based on big data analysis
Through a financial risk control system based on big data analysis, the NLP and LSTM models combined with improved particle swarm algorithms can collect and analyze financial data in real time, identify risks and generate prevention and control measures, and solve the problem of traditional risk control methods when facing massive data. Identifying problems of risk lag and inaccuracy has achieved effective risk management in the financial market.
Patent Information
- Application Number
- CN202411046324.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-01
- Publication Date
- 2024-11-26
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional financial risk control methods are difficult to quickly and accurately identify potential risks when facing massive data and complex transaction models, resulting in financial institutions that may suffer huge losses and threaten the stability of the financial market.
A financial risk control system based on big data analysis was designed, including financial data collection, data feature extraction, risk model establishment and financial risk prevention and control modules, using NLP natural language processing, LSTM long and short-term memory network and improved particle swarm algorithm IPSO optimization model, collects and analyzes financial data in real time, extracts risk characteristics, establishes a risk identification model, and generates prevention and control measures based on the risk level.
It realizes rapid processing and analysis of massive financial data, accurately judges risk conditions, provides scientific decision-making basis, can predict future risk trends and formulate response measures in advance to reduce financial risks.
Smart Images

Figure HDA0004973580930000011 
Figure HDA0004973580930000021 
Figure HDA0004973580930000022
Abstract
Description
Technical Field
[0001] The present invention relates to the field of financial risk control, and in particular to a financial risk control system based on big data analysis. Background Art
[0002] Under the wave of Internet finance, the financial industry is undergoing unprecedented changes. With the continuous advancement of technology and the increasing openness of the market, the volume of financial transactions has shown explosive growth, which has not only brought huge profits to financial institutions, but also brought unprecedented challenges. The diversification and complexity of financial risks have made traditional financial risk control methods seem incapable. Traditional financial risk control methods, such as manual review and qualitative analysis, are often difficult to quickly and accurately identify potential risks when faced with massive data and complex transaction patterns. This lag and inaccuracy may not only cause huge losses to financial institutions, but also threaten the stability of the entire financial market. Therefore, building an efficient and accurate financial risk control system has become a technical problem that needs to be solved in the era of Internet finance. Summary of the invention
[0003] The purpose of the present invention is to solve the above problems and to design a financial risk control system based on big data analysis.
[0004] To achieve the above objective, the technical solution of the present invention is that, further, in the above financial risk control system based on big data analysis, the financial risk control system includes the following steps:
[0005] A financial data acquisition module is used to obtain real-time financial data in the financial system and perform data preprocessing on the real-time financial data to obtain initial financial data;
[0006] A data feature extraction module, used to extract financial risk features from the initial financial data using an NLP natural language processing model to obtain feature financial data;
[0007] A risk model building module is used to build an LSTM financial risk identification model based on an LSTM long short-term memory network, and optimize the parameters in the LSTM financial risk identification model through an improved particle swarm algorithm IPSO to obtain an IPSO-LSTM financial risk identification model;
[0008] A financial risk identification module, used for inputting the characteristic financial data into the IPSO-LSTM financial risk identification model for training to obtain financial risk results;
[0009] The financial risk prevention and control module is used to determine the financial risk level based on the financial risk results and generate financial risk prevention and control measures according to the financial risk level.
[0010] Furthermore, in the above-mentioned financial risk control system based on big data analysis, the financial data collection module includes the following submodules:
[0011] The collection submodule is used to connect with the financial system through the API interface, monitor the data flow in the financial system in real time, implement the anomaly detection mechanism, mark the abnormal values in the data, and obtain marked financial data. The marked financial data at least includes bank transaction records, securities transaction records, social media behavior, public market information, and third-party credit data:
[0012] A filling submodule, used to delete outliers and fill missing values in the marked financial data to obtain complete financial data;
[0013] The normalization submodule is used to perform data normalization processing on the complete financial data to obtain normalized financial data, and to perform data segmentation or sampling on the normalized financial data to obtain initial financial data.
[0014] Furthermore, in the above-mentioned financial risk control system based on big data analysis, the data feature extraction module includes the following submodules:
[0015] A preprocessing submodule, used to perform text preprocessing on the initial financial data, wherein the text preprocessing includes word segmentation, stop word removal, stem extraction or word form restoration, etc., to obtain text financial data;
[0016] An identification submodule, for identifying the sentiment tendency in the text financial data by using an NLP model, where the sentiment tendency includes at least positive, negative, and neutral, reflects the market sentiment, and obtains the sentiment financial data;
[0017] An extraction submodule, used to extract words and phrases related to financial risks in the emotional financial data, including at least default, decline, and credit rating downgrade, to obtain text emotional financial data;
[0018] The conversion submodule is used to convert the text emotional financial data into numerical features by using Word2Vec word embedding technology to obtain feature financial data.
[0019] Furthermore, in the above-mentioned financial risk control system based on big data analysis, the risk model building module includes the following submodules:
[0020] A submodule is introduced to replace the random initialization particle population position in PSO based on PSO particle swarm algorithm;
[0021] An adjustment submodule, used for adjusting the inertia weight in the PSO particle swarm algorithm by using a nonlinear weight adjustment strategy;
[0022] The optimization submodule is used to optimize the learning factor in the PSO particle swarm algorithm, emphasizing the self-learning ability of particles in the early stage of the search and the social learning ability of particles in the later stage;
[0023] The calculation submodule is used to modify the fitness function to the average error of the model predicted position, calculate the distance from the predicted point to each real path, first determine whether the predicted point is on a horizontal path or a vertical path, and then calculate the distance from the point to the path, and use this distance value as the error value to obtain the improved particle swarm algorithm IPSO.
[0024] Furthermore, in the above-mentioned financial risk control system based on big data analysis, the risk model building module includes the following submodules:
[0025] The processing submodule is used to establish an LSTM financial risk identification model based on the LSTM long short-term memory network, perform denormalization on the output results of the fully connected layer in the model, and restore the results to the original data range;
[0026] The calculation submodule is used to calculate the absolute error according to the actual route of the moving target and the predicted result, optimize the hyperparameters of the LSTM financial risk identification model by using the improved particle swarm algorithm IPSO, and initialize the population position by using the Tent chaos method after setting the initial parameters of the algorithm;
[0027] The adjustment submodule is used to adjust the motion state of the particle through the speed and position update formula, update the inertia weight using the nonlinear inertia weight reduction strategy, and adjust the learning factor in combination with the asymmetric optimization learning factor strategy;
[0028] The submodule is obtained to calculate the modified fitness. The individual optimal and group optimal values are updated according to the fitness value, and the steps are executed repeatedly until 200 iterations are reached to obtain the financial risk results.
[0029] Furthermore, in the above-mentioned financial risk control system based on big data analysis, the financial risk prevention and control module includes the following submodules:
[0030] The judgment submodule is used to judge the financial risk level based on the financial risk results and generate financial risk prevention and control measures according to the financial risk level.
[0031] The standard submodule is used to set the risk level standard and map the financial risk results to low, medium and high risk levels;
[0032] An early warning submodule is used to trigger an early warning notification to relevant personnel if the financial risk level is a medium risk level or a high risk level;
[0033] The strategy submodule is used to formulate strategies for medium risk level and high risk level, and the strategies at least include increasing margin, adjusting investment portfolio, and strengthening auditing.
[0034] Furthermore, in the method for implementing the above-mentioned financial risk control system based on big data analysis, the method includes the following steps:
[0035] Acquire real-time financial data in a financial system, and perform data preprocessing on the real-time financial data to obtain initial financial data;
[0036] Extracting financial risk features from the initial financial data using an NLP natural language processing model to obtain characteristic financial data;
[0037] An LSTM financial risk identification model is established based on the LSTM long short-term memory network, and the parameters in the LSTM financial risk identification model are optimized by using the improved particle swarm algorithm IPSO to obtain the IPSO-LSTM financial risk identification model;
[0038] Input the characteristic financial data into the IPSO-LSTM financial risk identification model for training to obtain financial risk results;
[0039] The financial risk level is determined based on the financial risk results, and financial risk prevention and control measures are generated according to the financial risk level.
[0040] Furthermore, in the method for implementing the above-mentioned financial risk control system based on big data analysis, the method includes the following steps:
[0041] Connect with the financial system through the API interface, monitor the data flow in the financial system in real time, implement anomaly detection mechanism, mark the abnormal values in the data, and obtain marked financial data, which at least includes bank transaction records, securities transaction records, social media behavior, public market information, and third-party credit data:
[0042] Deleting outliers and filling missing values in the marked financial data to obtain complete financial data;
[0043] The complete financial data is subjected to data normalization processing to obtain normalized financial data, and the normalized financial data is subjected to data segmentation or sampling to obtain initial financial data.
[0044] Furthermore, in the method for implementing the above-mentioned financial risk control system based on big data analysis, the method includes the following steps:
[0045] Performing text preprocessing on the initial financial data, wherein the text preprocessing includes word segmentation, stop word removal, stem extraction or word form restoration, etc., to obtain text financial data;
[0046] Using an NLP model to identify sentiment trends in the textual financial data, where the sentiment trends include at least positive, negative, and neutral, reflecting market sentiment, and obtaining sentiment financial data;
[0047] Extracting words and phrases related to financial risks from the emotional financial data, including at least default, decline, and credit rating downgrade, to obtain text emotional financial data;
[0048] The text sentiment financial data is converted into numerical features using Word2Vec word embedding technology to obtain feature financial data.
[0049] Its beneficial effects are that by acquiring real-time financial data in the financial system, the real-time financial data is preprocessed to obtain initial financial data; the financial risk features in the initial financial data are extracted using the NLP natural language processing model to obtain characteristic financial data; an LSTM financial risk identification model is established based on the LSTM long short-term memory network, and the parameters in the LSTM financial risk identification model are optimized by the improved particle swarm algorithm IPSO to obtain the IPSO-LSTM financial risk identification model; the characteristic financial data is input into the IPSO-LSTM financial risk identification model for training to obtain financial risk results; the financial risk level is judged based on the financial risk results, and financial risk prevention and control measures are generated according to the financial risk level. Various types of financial data can be collected, processed and analyzed in real time, potential risk points can be found, and corresponding measures can be taken to prevent and respond. By using big data analysis technology, financial institutions can realize rapid processing and analysis of massive data, discover laws and anomalies in the data, and accurately judge the risk status. At the same time, big data analysis technology can also quantitatively evaluate risks and provide financial institutions with a more scientific basis for decision-making. By learning and simulating the human decision-making process, we can predict future risk trends based on historical data and current market conditions, and formulate corresponding response measures in advance. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the following detailed description of the preferred embodiment.The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting the invention.
[0051] Figure 1 This is a schematic diagram of a first embodiment of a financial risk control system based on big data analysis in an embodiment of the present invention;
[0052] Figure 2 This is a schematic diagram of a second embodiment of a financial risk control system based on big data analysis in an embodiment of the present invention;
[0053] Figure 3 This is a schematic diagram of a third embodiment of a financial risk control system based on big data analysis in an embodiment of the present invention; DETAILED DESCRIPTION
[0054] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0055] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "said" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.
[0056] The present invention will be described in detail below in conjunction with the accompanying drawings. Figure 1 As shown, a financial risk control system based on big data analysis includes the following modules:
[0057] The financial data acquisition module is used to obtain real-time financial data in the financial system, perform data preprocessing on the real-time financial data, and obtain initial financial data;
[0058] Specifically, the present embodiment also includes a collection submodule, which is used to connect with the financial system through an API interface, monitor the data flow in the financial system in real time, implement an anomaly detection mechanism, mark the outliers in the data, and obtain marked financial data, where the marked financial data at least includes bank transaction records, securities transaction records, social media behaviors, public market information, and third-party credit data; a filling submodule, which is used to delete outliers and fill missing values in the marked financial data to obtain complete financial data; and a normalization submodule, which is used to perform data normalization processing on the complete financial data to obtain normalized financial data, and perform data segmentation or sampling on the normalized financial data to obtain initial financial data.
[0059] The data feature extraction module is used to extract the financial risk features in the initial financial data using the NLP natural language processing model to obtain characteristic financial data;
[0060] Specifically, the present embodiment also includes a preprocessing submodule, which is used to perform text preprocessing on the initial financial data, the text preprocessing includes word segmentation, stop word removal, stem extraction or word form restoration, etc., to obtain text financial data; an identification submodule, which is used to use the NLP model to identify the sentiment tendency in the text financial data, the sentiment tendency includes at least positive, negative, and neutral, reflecting the market sentiment, and obtain emotional financial data; an extraction submodule, which is used to extract words and phrases related to financial risks in the emotional financial data, including at least default, decline, and credit rating downgrade, to obtain text emotional financial data; a conversion submodule, which is used to use the Word2Vec word embedding technology to convert the text emotional financial data into numerical features to obtain feature financial data.
[0061] The risk model building module is used to build an LSTM financial risk identification model based on the LSTM long short-term memory network, and optimize the parameters in the LSTM financial risk identification model through the improved particle swarm algorithm IPSO to obtain the IPSO-LSTM financial risk identification model;
[0062] Specifically, the present embodiment also includes an introduction submodule, which is used to introduce chaotic sequence initialization to replace the randomly initialized particle population position in the PSO based on the PSO particle swarm algorithm; an adjustment submodule, which is used to adjust the inertia weight in the PSO particle swarm algorithm using a nonlinear adjustment weight strategy; an optimization submodule, which is used to optimize the learning factor in the PSO particle swarm algorithm, emphasizing the self-learning ability of the particles in the early stage of the search and the social learning ability of the particles in the later stage; a calculation submodule, which is used to modify the fitness function to be the average error of the model predicted position, calculate the distance from the predicted point to each real path, first determine whether the predicted point is on a horizontal path or a vertical path, and then calculate the distance from the point to the path, and use this distance value as the error value to obtain an improved particle swarm algorithm IPSO.
[0063] Specifically, in this embodiment, the PSO particle swarm algorithm (PSO for short) is an optimization algorithm based on swarm intelligence, proposed by Kennedy and Eberhart in 1995. The algorithm simulates the foraging behavior of a flock of birds and finds the optimal solution to the problem through collaboration and information sharing between particles. The following is a detailed explanation of the PSO particle swarm algorithm:
[0064] 1. Algorithm Basics
[0065] Model setting: Compare the search space for solving the problem to the flight space of birds. Abstract each bird into a particle without mass and volume, representing a possible solution to the problem. The position of the particle represents the solution to the problem, and the speed of the particle represents the direction of change of the solution.
[0066] Optimization process: Each particle in the particle swarm adjusts its speed and position according to its own "flight experience" (historical best position pbest) and the group's "flight experience" (global best position gbest). By iteratively updating the speed and position of the particles, the optimal solution of the problem is gradually approached.
[0067] 2. Algorithm Process
[0068] Initialization: Randomly generate the initial position and velocity of the particles. Set the algorithm parameters, such as learning factors c1 and c2, inertia weight w, maximum velocity Vmax, maximum number of iterations Gk, etc.
[0069] Fitness evaluation: Calculate the fitness value of each particle, usually calculated by the objective function. Update individual optimal and global optimal: If the fitness value of the current particle is better than the fitness value of its historical best position pbest, update pbest. If the fitness value of the current particle is better than the fitness value of the global best position gbest, update gbest.
[0070] Update speed and position: Update the speed and position of the particle according to the formula. The speed update formula usually includes three parts: inertia term, individual cognition term and group cognition term. Limit the speed to ensure that it does not exceed the maximum speed Vmax.
[0071] Iteration termination condition: If the maximum number of iterations Gk is reached or other termination conditions are met (such as the fitness value reaches the preset threshold), the algorithm terminates. Output the global optimal solution gbest and its corresponding fitness value.
[0072] 3. Algorithm Features
[0073] Parallel search: Particle swarm algorithm is an efficient parallel search algorithm that can search multiple solution spaces simultaneously. Biosocial background: It has a good biosocial background and is easy to understand. Few parameters and easy to implement: It is easy to implement due to the small number of parameters, and has strong global search capabilities for nonlinear and multi-peak problems. Memory: The unique memory of particle swarm algorithm enables it to dynamically track the current search situation and adjust its search strategy. Insensitive to population size: Even when the population size decreases, the performance degradation is not very large.
[0074] The financial risk identification module is used to input characteristic financial data into the IPSO-LSTM financial risk identification model for training to obtain financial risk results;
[0075] Specifically, the present embodiment also includes a processing submodule, which is used to establish an LSTM financial risk identification model based on an LSTM long short-term memory network, and to perform a denormalization process on the output of the fully connected layer in the model to restore the result to the original data range; a calculation submodule, which is used to calculate the absolute error based on the actual route of the moving target and the predicted result, and to optimize the hyperparameters of the LSTM financial risk identification model using the improved particle swarm algorithm IPSO, and to initialize the population position using the Tent chaos method after setting the initial parameters of the algorithm; an adjustment submodule, which is used to adjust the motion state of the particles through the speed and position update formula, and to update the inertia weight using a nonlinear inertia weight reduction strategy, and to adjust the learning factor in combination with an asymmetric optimization learning factor strategy; and an acquisition submodule, which is used to calculate the modified fitness. The individual optimal and group optimal values are updated according to the fitness value, and the steps are executed repeatedly until 200 iterations are reached to obtain the financial risk result.
[0076] The financial risk prevention and control module is used to judge the financial risk level based on the financial risk results and generate financial risk prevention and control measures according to the financial risk level.
[0077] Specifically, the present embodiment also includes a judgment submodule, which is used to judge the financial risk level based on the financial risk results, and generate a financial risk prevention and control measures standard submodule according to the financial risk level, which is used to set the risk level standard and map it to low, medium and high risk levels according to the financial risk results; an early warning submodule, which is used to trigger an early warning notification to relevant personnel if the financial risk level is a medium risk level or a high risk level; and a strategy submodule, which is used to formulate strategies for medium risk levels and high risk levels, and the strategies at least include increasing margin, adjusting investment portfolios, and strengthening audits.
[0078] Its beneficial effects are that by obtaining real-time financial data in the financial system, the real-time financial data is preprocessed to obtain initial financial data; the financial risk features in the initial financial data are extracted using the NLP natural language processing model to obtain characteristic financial data; an LSTM financial risk identification model is established based on the LSTM long short-term memory network, and the parameters in the LSTM financial risk identification model are optimized by the improved particle swarm algorithm IPSO to obtain the IPSO-LSTM financial risk identification model; the characteristic financial data is input into the IPSO-LSTM financial risk identification model for training to obtain financial risk results; the financial risk level is judged based on the financial risk results, and financial risk prevention and control measures are generated according to the financial risk level. Various types of financial data can be collected, processed and analyzed in real time, potential risk points can be found, and corresponding measures can be taken to prevent and respond. By using big data analysis technology, financial institutions can realize the rapid processing and analysis of massive data, discover the laws and anomalies in the data, and accurately judge the risk status. At the same time, big data analysis technology can also quantitatively evaluate risks and provide financial institutions with a more scientific basis for decision-making. By learning and simulating the human decision-making process, we can predict future risk trends based on historical data and current market conditions, and formulate corresponding response measures in advance.
[0079] See also Figure 2 ,In a financial risk control system based on big data analysis, the data feature extraction module includes the following modules:
[0080] The preprocessing submodule is used to perform text preprocessing on the initial financial data. The text preprocessing includes word segmentation, stop word removal, stem extraction or word form restoration, etc., to obtain text financial data;
[0081] The identification submodule is used to use the NLP model to identify the sentiment tendency in text financial data. The sentiment tendency includes at least positive, negative, and neutral, reflecting the market sentiment and obtaining emotional financial data.
[0082] The extraction submodule is used to extract words and phrases related to financial risks in the emotional financial data, including at least default, decline, and credit rating downgrade, to obtain text emotional financial data;
[0083] The conversion submodule is used to convert text emotional financial data into numerical features using Word2Vec word embedding technology to obtain feature financial data.
[0084] See also Figure 3 ,In a spatiotemporal geographic information storage management service distribution system, the risk model building module includes the following modules:
[0085] The processing submodule is used to establish an LSTM financial risk identification model based on the LSTM long short-term memory network, perform denormalization on the output results of the fully connected layer in the model, and restore the results to the original data range;
[0086] The calculation submodule is used to calculate the absolute error based on the actual route of the moving target and the predicted result, and use the improved particle swarm algorithm IPSO to optimize the hyperparameters of the LSTM financial risk identification model. After setting the initial parameters of the algorithm, the Tent chaos method is used to initialize the population position;
[0087] The adjustment submodule is used to adjust the motion state of the particle through the speed and position update formula, update the inertia weight using the nonlinear inertia weight reduction strategy, and adjust the learning factor in combination with the asymmetric optimization learning factor strategy;
[0088] The submodule is obtained to calculate the modified fitness. The individual optimal and group optimal values are updated according to the fitness value, and the steps are executed repeatedly until 200 iterations are reached to obtain the financial risk results.
[0089] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.
Claims
1. A financial risk control system based on big data analysis, It is characterized in that The financial risk control system includes the following modules: A financial data acquisition module is used to obtain real-time financial data in the financial system and perform data preprocessing on the real-time financial data to obtain initial financial data; A data feature extraction module, used to extract financial risk features from the initial financial data using an NLP natural language processing model to obtain feature financial data; A risk model building module is used to build an LSTM financial risk identification model based on an LSTM long short-term memory network, and optimize the parameters in the LSTM financial risk identification model through an improved particle swarm algorithm IPSO to obtain an IPSO-LSTM financial risk identification model; A financial risk identification module, used for inputting the characteristic financial data into the IPSO-LSTM financial risk identification model for training to obtain financial risk results; The financial risk prevention and control module is used to determine the financial risk level based on the financial risk results and generate financial risk prevention and control measures according to the financial risk level.
2. A financial risk control system based on big data analysis as claimed in claim 1, It is characterized in that The financial data collection module includes the following submodules: The collection submodule is used to connect with the financial system through the API interface, monitor the data flow in the financial system in real time, implement the anomaly detection mechanism, mark the abnormal values in the data, and obtain marked financial data. The marked financial data at least includes bank transaction records, securities transaction records, social media behavior, public market information, and third-party credit data: A filling submodule, used for deleting outliers and filling missing values in the marked financial data to obtain complete financial data; The normalization submodule is used to perform data normalization processing on the complete financial data to obtain normalized financial data, and to perform data segmentation or sampling on the normalized financial data to obtain initial financial data.
3. A financial risk control system based on big data analysis as claimed in claim 1, It is characterized in that The data feature extraction module includes the following submodules: A preprocessing submodule, used to perform text preprocessing on the initial financial data, wherein the text preprocessing includes word segmentation, stop word removal, stem extraction or word form restoration, etc., to obtain text financial data; An identification submodule, for identifying the sentiment tendency in the text financial data by using an NLP model, where the sentiment tendency includes at least positive, negative, and neutral, reflects the market sentiment, and obtains the sentiment financial data; An extraction submodule, used to extract words and phrases related to financial risks in the emotional financial data, including at least default, decline, and credit rating downgrade, to obtain text emotional financial data; The conversion submodule is used to convert the text emotional financial data into numerical features by using Word2Vec word embedding technology to obtain feature financial data.
4. A financial risk control system based on big data analysis as claimed in claim 1, It is characterized in that The risk model building module includes the following submodules: A submodule is introduced to replace the random initialization particle population position in PSO based on PSO particle swarm algorithm; An adjustment submodule, used for adjusting the inertia weight in the PSO particle swarm algorithm by using a nonlinear weight adjustment strategy; The optimization submodule is used to optimize the learning factor in the PSO particle swarm algorithm, emphasizing the self-learning ability of particles in the early stage of the search and the social learning ability of particles in the later stage; The calculation submodule is used to modify the fitness function to the average error of the model predicted position, calculate the distance from the predicted point to each real path, first determine whether the predicted point is on a horizontal path or a vertical path, and then calculate the distance from the point to the path, and use this distance value as the error value to obtain the improved particle swarm algorithm IPSO.
5. A financial risk control system based on big data analysis as claimed in claim 1, It is characterized in that The risk model building module includes the following submodules: The processing submodule is used to establish an LSTM financial risk identification model based on the LSTM long short-term memory network, perform denormalization on the output results of the fully connected layer in the model, and restore the results to the original data range; The calculation submodule is used to calculate the absolute error according to the actual route of the moving target and the predicted result, optimize the hyperparameters of the LSTM financial risk identification model by using the improved particle swarm algorithm IPSO, and initialize the population position by using the Tent chaos method after setting the initial parameters of the algorithm; The adjustment submodule is used to adjust the motion state of the particle through the speed and position update formula, update the inertia weight using the nonlinear inertia weight reduction strategy, and adjust the learning factor in combination with the asymmetric optimization learning factor strategy; The submodule is obtained to calculate the modified fitness. The individual optimal and group optimal values are updated according to the fitness value, and the steps are executed repeatedly until 200 iterations are reached to obtain the financial risk results.
6. A financial risk control system based on big data analysis as claimed in claim 1, It is characterized in that The financial risk prevention and control module includes the following submodules: The judgment submodule is used to judge the financial risk level based on the financial risk results and generate financial risk prevention and control measures according to the financial risk level. The standard submodule is used to set the risk level standard and map the financial risk results to low, medium and high risk levels; An early warning submodule is used to trigger an early warning notification to relevant personnel if the financial risk level is a medium risk level or a high risk level; The strategy submodule is used to formulate strategies for medium risk level and high risk level, and the strategies at least include increasing margin, adjusting investment portfolio, and strengthening auditing.
7. A method for implementing a financial risk control system based on big data analysis as claimed in claim 1, It is characterized in that The method comprises the following steps: Acquire real-time financial data in a financial system, and perform data preprocessing on the real-time financial data to obtain initial financial data; Extracting financial risk features from the initial financial data using an NLP natural language processing model to obtain characteristic financial data; An LSTM financial risk identification model is established based on the LSTM long short-term memory network, and the parameters in the LSTM financial risk identification model are optimized by using the improved particle swarm algorithm IPSO to obtain the IPSO-LSTM financial risk identification model; Input the characteristic financial data into the IPSO-LSTM financial risk identification model for training to obtain financial risk results; The financial risk level is determined based on the financial risk results, and financial risk prevention and control measures are generated according to the financial risk level.
8. A method for implementing a financial risk control system based on big data analysis as claimed in claim 1, It is characterized in that The method comprises the following steps: Connect with the financial system through the API interface, monitor the data flow in the financial system in real time, implement anomaly detection mechanism, mark the abnormal values in the data, and obtain marked financial data, which at least includes bank transaction records, securities transaction records, social media behavior, public market information, and third-party credit data: Deleting outliers and filling missing values in the marked financial data to obtain complete financial data; The complete financial data is subjected to data normalization processing to obtain normalized financial data, and the normalized financial data is subjected to data segmentation or sampling to obtain initial financial data.
9. A method for implementing a financial risk control system based on big data analysis as claimed in claim 1, It is characterized in that The method comprises the following steps: Performing text preprocessing on the initial financial data, wherein the text preprocessing includes word segmentation, stop word removal, stem extraction or word form restoration, etc., to obtain text financial data; Using an NLP model to identify sentiment trends in the textual financial data, where the sentiment trends include at least positive, negative, and neutral, reflecting market sentiment, and obtaining sentiment financial data; Extracting words and phrases related to financial risks from the emotional financial data, including at least default, decline, and credit rating downgrade, to obtain text emotional financial data; The text sentiment financial data is converted into numerical features using Word2Vec word embedding technology to obtain feature financial data.
Citation Information
Cited By
Data risk control model construction method based on neural network
CN119417603A
Market supervision dynamic monitoring and intelligent risk early warning system and method based on big data
CN120013234A
