Financial risk intelligent identification and prediction method and system based on deep learning

Abstract

The invention relates to the technical field of deep learning data processing, and discloses a financial risk intelligent identification and prediction method and system based on deep learning, and the system comprises a data collection module and an intelligent identification module. According to the financial risk intelligent identification and prediction method and system based on deep learning, a data collection module collects a data set, an intelligent identification module extracts signal data influencing the financial market condition through an NLP technology, and an emotion feature data set is obtained through analysis; the intelligent identification module extracts financial market historical transaction data from morning to evening according to a sampling period, a risk prediction model is obtained through analysis of a GAN algorithm model, variability of a market structure is flexibly adapted, the deep learning identification capability is high, and the robustness is high. And the risk degree of the future financial market transaction is judged, the financial market supervisor is correspondingly prompted, the financial market supervisor is prompted to implement management measures, the risk management strategy is flexibly adjusted, and the prediction accuracy is high.

Description

Technical Field

[0001] The present invention relates to the technical field of deep learning data processing, and specifically to a method and system for intelligent identification and prediction of financial risks based on deep learning. Background Art

[0002] Financial risk refers to the possible losses that investors, financial institutions and financial markets may suffer due to various uncertainties in the financial market. The generation and spread of financial risks have an important impact on the stability and development of the financial market. There are many types of financial risks, mainly including market risk, credit risk, liquidity risk, operational risk, legal risk, etc. These risks are interrelated and influence each other, and together constitute the risk system of the financial market. Market risk refers to the risk of investment losses due to market price fluctuations. Market price fluctuations include changes in the prices of various financial assets such as interest rates, exchange rates, stock prices, and commodity prices. Market risk is the most common and direct risk in the financial market, and it has a direct impact on investors' returns and the balance sheet of financial institutions. In order to effectively manage market risks, investors and financial institutions need to use various financial tools and technologies for risk management, such as derivatives such as futures, options, swaps, and risk measurement methods such as value at risk VaR. Financial risk is an inherent attribute of the financial market, and it is closely related to the development and innovation of the financial market. Financial market participants need to fully recognize and understand various financial risks, use effective risk management methods and technologies, reduce the impact of financial risks on themselves, and contribute to the healthy development of the financial market.

[0003] At present, traditional financial risk intelligent identification and prediction methods train statistical models through historical price fluctuations, market trends and macroeconomic indicators, and then predict future market risks. Due to excessive reliance on historical data and the strong subjectivity of model hyperparameter settings, statistical models are difficult to adapt to the variability of financial market structures. Financial risk prediction has limitations, and accuracy and efficiency cannot be guaranteed. Summary of the invention

[0004] 1. Technical issues to be resolved

[0005] In view of the shortcomings of the prior art, the present invention provides a method and system for intelligent identification and prediction of financial risks based on deep learning, which has the advantages of strong deep learning identification ability, high robustness, flexible adjustment of risk management strategy and high prediction accuracy. It solves the problem that the statistical model hyperparameter setting of traditional financial risk intelligent identification and prediction methods is highly subjective, and the accuracy and efficiency of financial risk prediction cannot be guaranteed.

[0006] (II) Technical solution

[0007] To achieve the above object, the present invention provides the following technical solution: a method for intelligent identification and prediction of financial risks based on deep learning, comprising the following steps:

[0008] Step 1: The data acquisition module obtains the historical transaction data of the financial market at all time points through the network connection database and forms a historical data set;

[0009] Step 2: The data collection module obtains the financial statement data of all listed companies in the financial market through a network connection database and forms a financial report data set;

[0010] Step 3: The data collection module connects to the big data platform through the network to obtain the emotional signals reflected by all financial market participants in the trading process and composes them into a signal data set;

[0011] Step 4: The intelligent recognition module extracts the signal data that affects the financial market from the signal data set through NLP technology, and combines it with the financial report data set to analyze and obtain the emotional feature data set Qxtz;

[0012] Step 5: The intelligent recognition module is provided with a GAN algorithm model. The intelligent recognition module extracts part of the data in the historical data set, and substitutes the data into the GAN algorithm model in combination with the signal data set to analyze and obtain the risk prediction model Ycmx;

[0013] Step 6. The intelligent identification module is set with a risk threshold. The intelligent identification module substitutes the historical data set, financial report data set and signal data set into the risk prediction model Ycmx, analyzes and obtains the risk prediction results, and compares the risk prediction results with the risk threshold to determine the risk level of future financial market transactions and give corresponding prompts to financial market regulators.

[0014] Preferably, in step 1, the expression of the historical data set is {L1 d 、L2 d 、L3 d , ..., Ln d}, L1 d To Ln d The historical transaction data of the financial market corresponding to each time point in turn include transaction amount, transaction times and transaction volume. d represents the opening date of the financial market, and 1 to n represent that there are n groups of historical transaction data of the financial market.

[0015] Preferably, in step 2, the expression of the financial report data set is {C1 g 、C2 g 、C3 g , ..., Cb g}, C1 g To Cb gThey correspond to the financial statement data of each listed company in the financial market in turn, g represents the company name when the listed company is listed in the financial market, and 1 to b means there are b listed companies in the financial market.

[0016] Preferably, in step 3, the expression of the signal data set is {X1 q 、Z2 q 、X3 q ,...,Xh q}, X1 q To Xh q They correspond in turn to the emotional signals reflected by each financial market participant in the trading process. The emotional signals include news report data and social media data. q represents the group role of financial market participants, which includes investors, securities companies, listed companies and regulators. 1 to h means that there are h kinds of emotional signals in the financial market.

[0017] Preferably, in step 4, the calculation formula of the emotion feature data set Qxtz is as follows:

[0018] GS=∑ b (C1 g +C2 g +C3 g +...+Cb g )

[0019]

[0020] In the formula, Qxtz represents the emotion feature data set, ∑ b (C1 g +C2 g +C3 g +...+Cb g ) indicates that according to the financial report data set, the list of listed companies in the financial market GS is obtained. The list of listed companies in the financial market GS includes companies that are being listed, have been listed, and have been delisted. It means that 1 to h kinds of sentiment signals related to the list of listed companies in the financial market are extracted from the signal data set in sequence through NLP technology, which is the sentiment feature data set that affects the financial market situation.

[0021] Preferably, in step five, the intelligent identification module is provided with a sampling period of fixed duration, and is marked as GC, and the duration of a single sampling period GC is greater than or equal to seven days.

[0022] Preferably, in step 5, the intelligent recognition module extracts 5 groups of financial market historical transaction data within the sampling period from the historical data set in chronological order from early to late, and marks them as {GC 1 , GC 2 , GC3 , GC 4 , GC 5}, GC 1 Represents the historical trading data of the financial market within the first sampling period, GC 2 Represents the historical trading data of financial markets within the second sampling period, and later than GC 1 Opening date in GC 3 Represents historical financial market trading data within the third sampling period and later than GC 2 Opening date in GC 4 Represents historical financial market trading data within the fourth sampling period and later than GC 3 Opening date in GC 5 Represents historical financial market trading data within the fifth sampling period and later than GC 4 The opening date in.

[0023] Preferably, in step 5, the GAN algorithm model includes a generator G and a discriminator D, and is trained by a value function V(G, D), and the risk prediction model Ycmx calculation formula is as follows:

[0024] x=GC 1 +GC 2 +GC 3 +GC 4 +GC 5

[0025]

[0026] Ycmx=max G MIT D V(G, D)

[0027] In the formula, Ycmx represents the risk prediction model, x represents the historical transaction data of the financial market from the first to the fifth sampling period, that is, the real data, and Pdata(x) represents the real data distribution. Representing the latent space The distribution of D(x) represents the predicted probability of the discriminator D for the input real data x. Represents that the generator G converts the random vector The result of mapping to the data distribution is represents the expected value of correctly identifying from the distribution of the true data x, represents the expected value of data z sampled from the data distribution generated by the generator G being mistakenly identified as real data, max G min D V(G, D) represents the generator G and discriminator D of the optimal solution obtained during the training process of the GAN algorithm model, which is the risk prediction model.

[0028] Preferably, in step six, if the risk prediction result exceeds the risk threshold, it indicates that the volatility trend of the future financial market will change greatly, and the intelligent identification module determines that the risk level of future financial market transactions is high, prompting financial market regulators to implement management measures.

[0029] A financial risk intelligent identification and prediction system based on deep learning, including a database and a big data platform, the database is used to store and backup historical financial market transaction data and listed company financial statement data, the big data platform is used to collect news report data and social media data related to the financial market, and the financial risk intelligent identification and prediction system based on deep learning includes a data acquisition module and an intelligent identification module;

[0030] The data acquisition module includes a transaction data unit, a financial data unit and a market sentiment unit. The transaction data unit collects historical data sets through a network connection to a database, the financial data unit collects financial report data sets through a network connection to a database, and the market sentiment unit collects signal data sets through a network connection to a big data platform. The data acquisition module transmits the historical data sets, financial report data sets and signal data sets to the intelligent recognition module through the network;

[0031] The intelligent identification module includes a market analysis unit, a model training unit and a risk prediction unit. The market analysis unit extracts signal data that affects the financial market market from the signal data set through NLP technology, and combines it with the financial report data set to analyze and obtain the emotional feature data group Qxtz. The model training unit is provided with a sampling period of fixed duration, and extracts 5 groups of financial market historical transaction data from the historical data set from early to late according to the sampling period. The model training unit is provided with a GANK algorithm model, and the extracted financial market historical transaction data is combined with the signal data set and substituted into the GAN algorithm model to obtain the risk prediction model Ycmx through analysis. The risk prediction unit is provided with a risk threshold. The risk prediction unit substitutes the historical data set, the financial report data set and the signal data set into the risk prediction model Ycmx, analyzes and obtains the prediction result, and compares the risk prediction result with the risk threshold to judge the risk level of future financial market transactions and give corresponding prompts to financial market regulators.

[0032] Compared with the prior art, the present invention provides a method and system for intelligent identification and prediction of financial risks based on deep learning, which has the following beneficial effects:

[0033] 1. The present invention connects the database and the big data platform through a data acquisition module network, and classifies and collects historical data sets, financial report data sets and signal data sets and transmits them to the intelligent recognition module. The intelligent recognition module is provided with a market analysis unit, a model training unit and a risk prediction unit. The market analysis unit extracts the signal data that affects the financial market situation from the signal data set through the NLP technology, and combines the financial report data set to analyze and obtain the emotional feature data group Qxtz. The model training unit is provided with a sampling cycle of a fixed duration, and extracts 5 groups of financial market historical transaction data from the historical data set from early to late according to the sampling cycle, thereby ensuring the accuracy and prediction efficiency of the subsequent risk prediction model Ycmx. The model training unit is provided with a GANK algorithm model, and the extracted financial market historical transaction data is combined with the signal data set and substituted into the GAN algorithm model to obtain the risk prediction model Ycmx through analysis. The model hyperparameter setting is obtained based on the training of the financial market historical transaction data, and can flexibly adapt to the variability of the market structure, and has strong deep learning recognition ability and high robustness.

[0034] 2. The present invention sets a risk threshold through the risk prediction unit. The risk prediction unit substitutes the historical data set, financial report data set and signal data set into the risk prediction model Ycmx, analyzes and obtains the prediction results, and compares the risk prediction results with the risk threshold to capture all potential risk factors. The intelligent identification module compares the risk prediction results with the risk threshold. If the risk prediction results exceed the risk threshold, it means that the future financial market volatility trend will change greatly. The intelligent identification module determines that the risk level of future financial market transactions is high, and prompts financial market regulators to implement management measures and flexibly adjust risk management strategies with high prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a step diagram of the method of the present invention;

[0036] Figure 2 It is a schematic diagram of the system flow of the present invention. DETAILED DESCRIPTION

[0037] 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.

[0038] Traditional financial risk intelligent identification and prediction methods use historical price fluctuations, market trends and macroeconomic indicators to train statistical models and then predict future market risks. Due to excessive reliance on historical data and the strong subjectivity of model hyperparameter settings, statistical models are difficult to adapt to the variability of financial market structures. Financial risk prediction has limitations, and accuracy and efficiency cannot be guaranteed. Therefore, a financial risk intelligent identification and prediction method and system based on deep learning is provided. Please refer to Figure 1-Figure 2 , the method comprises the following steps:

[0039] Step 1: The data acquisition module obtains the historical transaction data of the financial market at all time points through the network connection database, and forms a historical data set. The expression of the historical data set is {L1 d 、L2 d 、L3 d , ..., Ln d}, L1 d To Ln d The historical transaction data of the financial market corresponding to each time point in turn, including transaction amount, transaction number and transaction volume. For example, the amount, transaction number, transaction volume and transaction object of each transaction from the opening to the closing of the exchange on a single day are counted. The data has strong traceability, which is conducive to the subsequent analysis of the macroeconomic status of the financial market. d represents the opening date of the financial market, and 1 to n represent that there are n groups of historical transaction data of the financial market;

[0040] Step 2: The data acquisition module obtains the financial statement data of all listed companies in the financial market through a network connection database, and forms a financial report data set. The expression of the financial report data set is {C1g, C2g, C3g, ..., Cbg}. C1g to Cbg correspond to the financial statement data of each listed company in the financial market in turn. These financial statement data can truly and accurately reflect the operating status of listed companies, safeguard the right of investors to know, and maintain healthy competition in the financial market is the key to ensure the healthy and stable development of the market. g represents the company name of the listed company when it is listed in the financial market, and 1 to b represent that there are b listed companies in the financial market;

[0041] Step 3: The data acquisition module connects to the big data platform through the network to obtain the emotional signals reflected by all financial market participants in the transaction process, and forms a signal data set. The expression of the signal data set is {X1 q , X2 q 、X3 q ,...,Xh q}, X1 q To Xh qIt corresponds to the emotional signals reflected by each financial market participant in the trading process. Emotional signals include news report data and social media data. Various emotional data will have an impact on market price trends and will gather sharply, thus triggering changes in market conditions. q represents the group role of financial market participants, including investors, securities companies, listed companies and regulators. 1 to h means that there are h kinds of emotional signals in the financial market.

[0042] Step 4: The intelligent recognition module uses NLP technology to extract the signal data that affects the financial market in the signal data set, and combines it with the financial report data set to analyze and obtain the emotional feature data set Qxtz. The calculation formula is as follows:

[0043] GS=∑ b (C1g+C2g+C3g+...+Cbg)

[0044]

[0045] In the formula, Qxtz represents the emotion feature data set, ∑ b (C1 g +C2 g +C3 g +...+Cbg) indicates that according to the financial report data set, the list of listed companies in the financial market GS is obtained. The list of listed companies in the financial market GS includes companies that are being listed, have been listed, and have been delisted. It means that 1 to h emotional signals related to the list of listed companies in the financial market are extracted from the signal data set in sequence through NLP technology, which is the emotional feature data set that affects the financial market situation, such as stock price adjustment news related to listed companies, investors' evaluation emotions towards listed companies, policy changes in the industry field where listed companies are located, and securities companies' evaluation reports on listed companies.

[0046] Step 5: The intelligent recognition module is provided with a GAN algorithm model. The intelligent recognition module extracts part of the data in the historical data set, and substitutes the data into the GAN algorithm model in combination with the signal data set to analyze and obtain the risk prediction model Ycmx;

[0047] The intelligent recognition module sets a fixed-length sampling cycle and marks it as GC. The duration of a single sampling cycle GC is greater than or equal to seven days. The intelligent recognition module extracts 5 groups of financial market historical transaction data within the sampling cycle from the historical data set in chronological order from early to late, and marks them as {Gc 1 , GC 2 , GC 3 , GC 4 , GC 5}, 5 sets of sampled data will be used as training data to substitute into the GAN algorithm model, GC 1Represents the historical trading data of the financial market within the first sampling period, GC 2 Represents the historical trading data of financial markets within the second sampling period, and later than GC 1 Opening date in GC 3 Represents historical financial market trading data within the third sampling period and later than GC 2 Opening date in GC 4 Represents historical financial market trading data within the fourth sampling period and later than GC 3 Opening date in GC 5 Represents historical financial market trading data within the fifth sampling period and later than GC 4 The opening date in the 5 sets of sampling data ensures the accuracy and prediction efficiency of the subsequent risk prediction model Ycmx. The GAN algorithm model includes a generator G and a discriminator D, and is trained through the value function V(G, D). The calculation formula of the risk prediction model Ycmx is as follows:

[0048] x=GC 1 +GC 2 +GC 3 +GC 4 +GC 5

[0049]

[0050] Ycmx=max G min D V(G, D)

[0051] In the formula, Ycmx represents the risk prediction model, x represents the historical transaction data of the financial market within the first to fifth sampling periods, that is, the real data. In the actual training process, the real data x can be set to the historical transaction data of the financial market within seven days at the shortest. According to the actual market status, the short-term prediction model or the long-term prediction model can be trained. Pdata(x) represents the real data distribution. Representing the latent space The distribution of D(x) represents the predicted probability of the discriminator D for the input real data x. Represents that the generator G converts the random vector The result of mapping to the data distribution is represents the expected value of correctly identifying from the distribution of the true data x, Represents data sampled from the data distribution generated by generator G The expected value of data that is mistakenly identified as real data, max G min DV(G, D) represents the generator G and discriminator D of the optimal solution obtained during the training of the GAN algorithm model, which is the risk prediction model. The model hyperparameter settings are obtained by training based on historical transaction data of the financial market. It can flexibly adapt to the variability of the market structure, and has strong deep learning recognition ability and high robustness.

[0052] Step 6. The intelligent recognition module is set with a risk threshold. The intelligent recognition module substitutes the historical data set, financial report data set and signal data set into the risk prediction model Ycmx, analyzes and obtains the risk prediction results, and captures all potential risk factors. The intelligent recognition module compares the risk prediction results with the risk threshold. If the risk prediction results exceed the risk threshold, it means that the future financial market volatility trend will change greatly. The intelligent recognition module determines that the risk level of future financial market transactions is high, and prompts financial market regulators to implement management measures and flexibly adjust risk management strategies with high prediction accuracy.

[0053] The above-mentioned financial risk intelligent identification and prediction method based on deep learning is implemented by relying on the financial risk intelligent identification and prediction system based on deep learning, including a database and a big data platform. The database is used to store and backup historical transaction data of the financial market and financial statement data of listed companies. The big data platform is used to collect news report data and social media data related to the financial market. The system includes a data collection module and an intelligent identification module.

[0054] The data collection module includes a transaction data unit, a financial data unit and a market sentiment unit. The transaction data unit collects historical data sets through a network connection to a database, the financial data unit collects financial report data sets through a network connection to a database, and the market sentiment unit collects signal data sets through a network connection to a big data platform. The data collection module transmits historical data sets, financial report data sets and signal data sets to the intelligent recognition module through the network;

[0055] The intelligent recognition module includes a market analysis unit, a model training unit and a risk prediction unit. The market analysis unit extracts signal data that affects the financial market from the signal data set through NLP technology, and combines it with the financial report data set to analyze and obtain the sentiment feature data group Qxtz. The model training unit is set with a sampling period of fixed duration, and extracts 5 groups of financial market historical transaction data from the historical data set from early to late according to the sampling period. The model training unit is set with a GANK algorithm model, and the extracted financial market historical transaction data is combined with the signal data set and substituted into the GAN algorithm model to obtain the risk prediction model Ycmx after analysis. The deep learning recognition ability is strong and the robustness is high. The risk prediction unit is set with a risk threshold. The risk prediction unit substitutes the historical data set, the financial report data set and the signal data set into the risk prediction model Ycmx, analyzes and obtains the prediction results, and compares the risk prediction results with the risk threshold to judge the risk level of future financial market transactions, and gives corresponding prompts to financial market regulators, so as to flexibly adjust the risk management strategy with high prediction accuracy.

[0056] 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. A method for intelligent identification and prediction of financial risks based on deep learning, characterized by: The following steps are involved: Step 1: The data acquisition module obtains the historical transaction data of the financial market at all time points through the network connection database and forms a historical data set; Step 2: The data collection module obtains the financial statement data of all listed companies in the financial market through a network connection database and forms a financial report data set; Step 3: The data collection module connects to the big data platform through the network to obtain the emotional signals reflected by all financial market participants in the trading process and composes them into a signal data set; Step 4: The intelligent recognition module extracts the signal data that affects the financial market from the signal data set through NLP technology, and combines it with the financial report data set to analyze and obtain the emotional feature data set Qxtz; Step 5: The intelligent recognition module is provided with a GAN algorithm model. The intelligent recognition module extracts part of the data in the historical data set, and substitutes the data into the GAN algorithm model in combination with the signal data set to analyze and obtain the risk prediction model Ycmx; Step 6. The intelligent identification module is set with a risk threshold. The intelligent identification module substitutes the historical data set, financial report data set and signal data set into the risk prediction model Ycmx, analyzes and obtains the risk prediction results, and compares the risk prediction results with the risk threshold to determine the risk level of future financial market transactions and give corresponding prompts to financial market regulators.

2. The method for intelligent identification and prediction of financial risks based on deep learning according to claim 1, characterized in that: In step 1, the expression of the historical data set is {L1 d 、L2 d 、L3 d , ..., Ln d }, L1 d To Ln d The historical transaction data of the financial market corresponding to each time point in turn include transaction amount, transaction times and transaction volume. d represents the opening date of the financial market, and 1 to n represent that there are n groups of historical transaction data of the financial market.

3. The method for intelligent identification and prediction of financial risks based on deep learning according to claim 2 is characterized in that: In the step 2, the expression of the financial report data set is {C1g, C2g, C3g, ..., Cbg}, where C1g to Cbg correspond to the financial statement data of each listed company in the financial market, g represents the company name when the listed company is listed in the financial market, and 1 to b represent that there are b listed companies in the financial market.

4. The method for intelligent identification and prediction of financial risks based on deep learning according to claim 3 is characterized by: In step 3, the expression of the signal data set is {X1 q , X2 q 、X3 q ,...,Xh q }, X1 q To Xh q They correspond in turn to the emotional signals reflected by each financial market participant in the trading process. The emotional signals include news report data and social media data. q represents the group role of financial market participants, which includes investors, securities companies, listed companies and regulators. 1 to h means that there are h kinds of emotional signals in the financial market.

5. The method for intelligent identification and prediction of financial risks based on deep learning according to claim 4 is characterized in that: In step 4, the calculation formula of the emotion feature data set Qxtz is as follows: GS=Σ b (C1 g +C2 g +C3 g +...+Cb g ) In the formula, Qxtz represents the emotion feature data set, ∑ b (C1 g +C2 g +C3 g +...+Cbg) indicates that according to the financial report data set, the list of listed companies in the financial market GS is obtained. The list of listed companies in the financial market GS includes companies that are being listed, have been listed, and have been delisted. It means that 1 to h kinds of sentiment signals related to the list of listed companies in the financial market are extracted from the signal data set in sequence through NLP technology, which is the sentiment feature data set that affects the financial market situation.

6. The method for intelligent identification and prediction of financial risks based on deep learning according to claim 5 is characterized by: In the step 5, the intelligent identification module is set with a sampling period of fixed duration and marked as GC, and the duration of a single sampling period GC is greater than or equal to seven days.

7. The method for intelligent identification and prediction of financial risks based on deep learning according to claim 6 is characterized by: In step 5, the intelligent recognition module extracts 5 groups of financial market historical transaction data within the sampling period from the historical data set in chronological order from early to late, and marks them as {GC 1 , GC 2 , GC 3 , GC 4 , GC 5 }, GC 1 Represents the historical trading data of the financial market within the first sampling period, GC 2 Represents the historical trading data of financial markets within the second sampling period, and later than GC 1 Opening date in GC 3 Represents historical financial market trading data within the third sampling period and later than GC 2 Opening date in GC 4 Represents historical financial market trading data within the fourth sampling period and later than GC 3 Opening date in GC 5 Represents historical financial market trading data within the fifth sampling period and later than GC 4 The opening date in.

8. The method for intelligent identification and prediction of financial risks based on deep learning according to claim 7 is characterized in that: In step 5, the GAN algorithm model includes a generator G and a discriminator D, and is trained by a value function V(G, D). The risk prediction model Ycmx is calculated as follows: x=GC 1 +GC 2 +GC 3 +GC 4 +GC 5 Ycmx=max G min D V(G,D) In the formula, Ycmx represents the risk prediction model, x represents the historical transaction data of the financial market from the first to the fifth sampling period, that is, the real data, and Pdata(x) represents the real data distribution. Representing the latent space The distribution of D(x) represents the predicted probability of the discriminator D for the input real data x. Represents that the generator G converts the random vector The result of mapping to the data distribution is represents the expected value of correctly identifying from the distribution of the true data x, Represents data sampled from the data distribution generated by generator G The expected value of data that is mistakenly identified as real data, max G min D V(G, D) represents the generator G and discriminator D of the optimal solution obtained during the training process of the GAN algorithm model, which is the risk prediction model.

9. The method for intelligent identification and prediction of financial risks based on deep learning according to claim 8, characterized in that: In step six, if the risk prediction result exceeds the risk threshold, it means that the future financial market volatility trend will change greatly. The intelligent identification module determines that the risk level of future financial market transactions is high, and prompts financial market regulators to implement management measures.

10. A financial risk intelligent identification and prediction system based on deep learning, applied to the financial risk intelligent identification and prediction method based on deep learning according to any one of claims 1 to 9, comprising a database and a big data platform, the database is used to store and back up historical financial market transaction data and listed company financial statement data, and the big data platform is used to collect news report data and social media data related to the financial market, characterized in that: The financial risk intelligent identification and prediction system based on deep learning includes a data acquisition module and an intelligent identification module; The data acquisition module includes a transaction data unit, a financial data unit and a market sentiment unit. The transaction data unit collects historical data sets through a network connection to a database, the financial data unit collects financial report data sets through a network connection to a database, and the market sentiment unit collects signal data sets through a network connection to a big data platform. The data acquisition module transmits the historical data sets, financial report data sets and signal data sets to the intelligent recognition module through the network; The intelligent identification module includes a market analysis unit, a model training unit and a risk prediction unit. The market analysis unit extracts signal data that affects the financial market market from the signal data set through NLP technology, and combines it with the financial report data set to analyze and obtain the emotional feature data group Qxtz. The model training unit is provided with a sampling period of fixed duration, and extracts 5 groups of financial market historical transaction data from the historical data set from early to late according to the sampling period. The model training unit is provided with a GANK algorithm model, and the extracted financial market historical transaction data is combined with the signal data set and substituted into the GAN algorithm model to obtain the risk prediction model Ycmx through analysis. The risk prediction unit is provided with a risk threshold. The risk prediction unit substitutes the historical data set, the financial report data set and the signal data set into the risk prediction model Ycmx, analyzes and obtains the prediction result, and compares the risk prediction result with the risk threshold to judge the risk level of future financial market transactions and give corresponding prompts to financial market regulators.

Images