Method for obtaining security market stationary strategy based on active disturbance rejection control

Abstract

The invention provides a method for obtaining a security market stationary strategy based on active-disturbance-rejection control, and the method comprises the steps: obtaining multi-source heterogeneous security market data through employing a crawler technology, carrying out the processing through employing a data warehouse, obtaining target data, and storing the target data in the data warehouse; analyzing the target data through a natural language processing method and an SKEP sentiment analysis model, and constructing a sentiment tendency index of the security market; constructing an ADRC model based on the target data and the emotional tendency index; according to the target data and the emotional tendency index, simulating security trading markets in various states based on a Duffing-Holms model, and respectively constructing corresponding virtual simulation models; and analyzing a stationary strategy of the virtual simulation model in various states by adopting an ADRC model to obtain a target stationary strategy. According to the invention, quantitative analysis can be carried out based on financial public opinions, and security transaction environments in various states are simulated to analyze the effectiveness of the stationary strategy, so that the stationary strategy of the security market is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to a method for obtaining a stable strategy for a securities market based on self-disturbance rejection control. Background Art

[0002] In the field of financial securities markets, public opinion data is usually only judged in a positive or negative text form, which cannot quantify the impact of public opinion on the securities market and is difficult to predict financial risks. In addition, since the securities market is affected by multiple factors and presents nonlinear characteristics, it is difficult for existing securities market analysis systems to accurately build a simulation environment. Usually, they can only build a single market trend stabilization strategy. The lack of comprehensive market analysis may lead to decision-making errors, and it is difficult to fully identify potential market risks and obtain effective securities market stabilization strategies.

[0003] Therefore, there is an urgent need for a securities market stabilization strategy method that can quantify financial public opinion and conduct a comprehensive analysis of securities trading situations under various conditions. Summary of the invention

[0004] Based on this, it is necessary to provide a method for obtaining a stable strategy for the securities market based on active anti-disturbance control to address the above technical problems.

[0005] A method for obtaining a securities market stabilization strategy based on active disturbance rejection control comprises the following steps: using crawler technology to obtain multi-source heterogeneous securities market data, and processing the securities market data through a data warehouse to obtain target data and store it in the data warehouse; using a natural language processing method and a SKEP sentiment analysis model to analyze the target data and construct a sentiment tendency index for the securities market; constructing an ADRC model based on the target data and the sentiment tendency index; simulating securities trading markets under various conditions based on a Duffing-Holms model according to the target data and the sentiment tendency index, and respectively constructing corresponding virtual simulation models; and using the ADRC model to analyze the stabilization strategies of the virtual simulation models under various conditions to obtain a target stabilization strategy.

[0006] In one of the embodiments, the crawler technology is used to obtain multi-source heterogeneous securities market data, and the securities market data is processed through a data warehouse to obtain target data and store it in the data warehouse, including: obtaining multi-source heterogeneous securities market data through a public database, actual securities market transaction data and crawler technology, and storing it in a data warehouse; synchronizing, performing data quality inspection, data cleaning and normalization, data aggregation, data conversion and data application on the securities market data through the data warehouse to obtain target data, and storing it in the data warehouse.

[0007] In one embodiment, the target data is analyzed by a natural language processing method and a SKEP sentiment analysis model to construct a sentiment index of the securities market, including: calculating the word frequency of the securities market public opinion in the target data by a natural language processing method, and the formula is:

[0008] TF(i,j)=n(i,j) / ∑K n (k,j) (1);

[0009] In the formula, n(i,j) represents the number of times word t appears in document j, ∑K n (k,j) represents the total number of all feature words in document j; the inverse document frequency of the securities market public opinion in the target data is calculated as follows:

[0010] IDF(i,j)=log(|D| / (1+D_t)) (2);

[0011] In the formula, |D| represents the total number of documents in the corpus, and D_t represents the number of documents containing word t; the TF-IDF value is calculated based on the word frequency and inverse document frequency, and the formula is:

[0012] TF-IDF(i,j)=TF(i,j)×IDF(i,j) (3);

[0013] Based on the SKEP sentiment analysis model, the divergence index, attention index and sentiment index are constructed; wherein the divergence index is used to measure the degree of divergence of opinions among investors, and the formula is:

[0014]

[0015] Where B is the number of bullish people and S is the number of bearish people. The calculation formula of the sentiment index is:

[0016] ISI = (V × σ) / S (5);

[0017] In the formula, V represents the trading volume, S represents the investor survey score, reflecting the investor's optimism or pessimism, and σ represents volatility. The calculation formula of the attention index is:

[0018] isi=(V t ×W)+(σ×W)+(else×W) (6);

[0019] Where V t It represents the rate of change of trading volume, and W represents the weight of the corresponding indicator.

[0020] In one embodiment, the ADRC model is constructed based on the target data and the sentiment index, including: constructing a tracking differentiator, an extended state observer and a nonlinear state error feedback control law; wherein the tracking differentiator is used to make the controlled object smoothly transition from the current value to the target value according to the difference between the actual market value and the target value, and the formula is:

[0021]

[0022] In the formula, v1(t) represents the transition process, is the desired stable target, γ is the undetermined parameter that determines the speed of the transition process. The larger the γ, the shorter the time it takes for the transition process to reach the target value. The extended state observer is used to obtain securities trend information, and the formula is:

[0023]

[0024] In the formula, z1(t) is the current information of the controlled object, z2(t) and z3(t) reflect the trend information of the controlled object, x1(t) represents the stock market price, α1, α2, δ1, δ2, β 01 ,β 02 ,β 03 , b0 is the ESO parameter to be determined, e(t) represents the error of the controlled system, and the fal function is:

[0025]

[0026] Wherein, α and δ are parameters related to the fal function; the nonlinear state error feedback control law formula is:

[0027]

[0028] Where u(t) is the securities market stabilization strategy generated by ADRC, u(t)>0 means buying securities, otherwise it means selling securities, β1, β2 are parameters to be selected, e1(t) represents the current error, and e2(t) represents the error change rate.

[0029] In one embodiment, according to the target data and the sentiment index, based on the Duffing-Holms model, the stock market under various conditions is simulated, and corresponding virtual simulation models are respectively constructed, including: according to the target data and the sentiment index, the Duffing-Holms model is obtained, which is:

[0030]

[0031] In the formula, x1(t) represents the stock market price, is the speed of change of the securities market, represents the acceleration of market changes; c represents the first policy parameter; d represents the second policy parameter; z is the speculative disturbance parameter; ω represents the self-regulation frequency of the securities market; a and b are the system rigidity coefficients; let The Duffing-Holms model of the transition from random walk state to chaotic market crisis state is:

[0032]

[0033] The Duffing-Holms model under the influence of external factors is:

[0034]

[0035] In the formula, J represents the jump process amplitude during the violent fluctuation of securities market prices, Q(t) represents the Poisson distribution jump process; σ represents the volatility of securities market prices, and B(t) represents the irregular Brownian motion process; the Duffing-Holms model under the stable strategy is:

[0036]

[0037] In the formula, u(t) represents a stable strategy, u(t)>0 represents the execution of a buy strategy in the market, and u(t)<0 represents the execution of a sell strategy in the market.

[0038] In one of the embodiments, the ADRC model is used to analyze the smooth strategy of the virtual simulation model under various states to obtain a target smooth strategy, including: obtaining simulation parameters of the ADRC model and bringing them into the ADRC model, wherein the simulation parameters are used to perform preliminary control on the controlled object; based on the virtual simulation model under various states, the ADRC model with the simulation parameters brought into is used to simulate the smooth strategy, and the effectiveness of the smooth strategy is judged, and the target smooth strategy is obtained according to the judgment result.

[0039] Compared with the prior art, the advantages and beneficial effects of the present invention are: using crawler technology to obtain multi-source heterogeneous securities market data, and processing the securities market data through a data warehouse to obtain target data and store it in the data warehouse, so as to achieve comprehensive and accurate data acquisition, reduce data processing costs and improve collection efficiency at the same time, analyze the target data through a natural language processing method and a SKEP sentiment analysis model, and construct a sentiment index of the securities market, thereby quantifying the securities market public opinion and investor sentiment index, so as to enhance the timeliness and information perception frontier of a stable strategy, and achieve accurate capture of changes in market sentiment, and timely identify risks; construct an ADRC model based on target data and the sentiment index, simulate securities trading markets under various conditions based on the Duffing-Holms model according to the target data and the sentiment index, respectively construct corresponding virtual simulation models, and achieve simulation of market trading environments under various conditions, so as to analyze the effectiveness of stable strategies under various conditions; use the ADRC model to analyze the stable strategies of virtual simulation models under various conditions to obtain target stable strategies, and evaluate the effectiveness of stable strategies by analyzing the trend of the securities market, thereby obtaining effective stable strategies and improving work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A flow chart of a method for obtaining a stable strategy for a securities market based on active disturbance rejection control in one embodiment;

[0041] Figure 2 A schematic diagram of the architecture of a method for obtaining a stable strategy for a securities market based on active disturbance rejection control in an embodiment;

[0042] Figure 3 An ARDC model in a securities market in one embodiment;

[0043] Figure 4 A schematic diagram of the changing trend and changing speed of the chaotic state of the securities market (excluding external disturbances) in one embodiment;

[0044] Figure 5 A schematic diagram of the changing trend and changing speed of the chaotic state (including external disturbance) of the securities market in one embodiment;

[0045] Figure 6 This is a diagram showing the effect of a stabilization strategy when the market fluctuates violently in an embodiment;

[0046] Figure 7 This is a diagram showing the effect of a stabilization strategy when the market soars in one embodiment;

[0047] Figure 8 This is a diagram showing the effect of a stabilization strategy when the market crashes in one embodiment;

[0048] Fig. 9 It is a diagram showing the effect of a stable strategy with different transition parameters in one embodiment;

[0049] Fig.10 is the stable cost of the securities market in one embodiment with or without the disturbance term. DETAILED DESCRIPTION

[0050] Before describing the specific embodiments of the present invention, the overall concept of the present invention is described as follows:

[0051] The present invention is mainly developed based on the securities market data analysis process. The current securities market analysis system is unable to quantify financial public opinion, lacks comprehensive market analysis, and is difficult to obtain an effective securities market stabilization strategy.

[0052] Therefore, the present invention proposes a method for obtaining a stable strategy for the securities market based on self-disturbance rejection control, which uses crawler technology to obtain multi-source heterogeneous securities market data, and processes the securities market data through a data warehouse to obtain target data and store it in the data warehouse, thereby achieving comprehensive and accurate data acquisition, which can reduce the data processing cost while improving the collection efficiency. The target data is analyzed through a natural language processing method and a SKEP sentiment analysis model, and a sentiment tendency index of the securities market is constructed, thereby quantifying the securities market public opinion and the investor sentiment index, so as to enhance the timeliness and information perception frontier of the stable strategy, and achieve accurate capture of market sentiment changes and timely identification of risks; an ADRC model is constructed based on the target data and the sentiment tendency index, and according to the target data and the sentiment tendency index, the securities trading market under various states is simulated based on the Duffing-Holms model, and corresponding virtual simulation models are respectively constructed to achieve simulation of the market trading environment under various states, so as to analyze the effectiveness of the stable strategy under various states; the ADRC model is used to analyze the stable strategy of the virtual simulation model under various states to obtain the target stable strategy, and the effectiveness of the stable strategy is evaluated by analyzing the trend of the securities market, thereby obtaining an effective stable strategy and improving work efficiency.

[0053] After introducing the overall concept of the present invention, in order to make the purpose, technical solution and advantages of the present invention more clear, the present invention is further described in detail by specific implementation methods in combination with the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0054] In one embodiment, Figure 1 As shown, a method for obtaining a stable strategy for the securities market based on active disturbance rejection control is provided, comprising the following steps:

[0055] Step S110, using crawler technology to obtain multi-source heterogeneous securities market data, and processing the securities market data through a data warehouse to obtain target data and store it in the data warehouse.

[0056] Specifically, in order to obtain relevant data on the securities market, public databases, commercial databases and python crawler technology are used to collect relevant data in the securities market, realize automatic collection and real-time updating of data, and avoid the data information island problem brought by a single data set.

[0057] Among them, step S110 includes: obtaining multi-source heterogeneous securities market data through public databases, actual securities market transaction data and crawler technology, and storing them in a data warehouse; synchronizing, testing data quality, cleaning and normalizing, aggregating, converting and applying securities market data through the data warehouse to obtain target data, and storing them in the data warehouse.

[0058] Specifically, multi-source heterogeneous data is obtained through public databases, actual transaction data of the securities market and crawler technology, and stored in the data warehouse. Through data warehouse synchronization, data warehouse OOS (Out-of-Specification) layer quality inspection, data warehouse DWD (Data Warehouse Details) layer data governance, data warehouse DWS (Data Warehouse Service, data warehouse service), DWT (Data Warehouse Transformation) layer data conversion, and data warehouse ADS (Application Data Service) layer output data, the target data is obtained and stored in the data warehouse again, so as to achieve the acquisition and preprocessing of diversified securities market data, so as to obtain more comprehensive and accurate data support, so as to facilitate the mining of deep connections and rules of data, reduce data processing costs, improve collection efficiency, and realize cross-source data mining.

[0059] Step S120, analyzing the target data through natural language processing methods and SKEP sentiment analysis model to construct a sentiment index of the securities market.

[0060] Specifically, after obtaining relevant data on the securities market, the natural language processing method and the SKEP (Sentiment Knowledge Enhanced Pre-training) sentiment analysis model are used to analyze the target data. By analyzing the language structure of the securities market, the text data is reconstructed, and the sentiment tendency index of the securities market is constructed. Indicators such as market public opinion and investor sentiment tendencies are effectively extracted and constructed from massive text data, so as to quantify the public opinion in the securities market and improve the sensitivity and cutting-edge nature of observing market changes.

[0061] Among them, step S120 includes: calculating the term frequency (TF) of the securities market public opinion in the target data by natural language processing method. TF is used to measure the frequency of a word appearing in a specific document. The formula is:

[0062] TF(i,j)=n(i,j) / ∑K n (k,j) (1);

[0063] In the formula, n(i,j) represents the number of times word t appears in document j, ∑K n (k, j) represents the total number of all feature words in document j; the inverse document frequency (IDF, InverseDocumentFrequency) of the securities market public opinion in the target data is calculated as follows:

[0064] IDF(i,j)=log(|D| / (1+D_t)) (2);

[0065] In the formula, |D| represents the total number of documents in the corpus, D_t represents the number of documents containing word t, and the denominator is used to prevent the denominator from being 0. The TF-IDF value is calculated based on the word frequency and inverse document frequency. The formula is:

[0066] TF-IDF(i,j)=TF(i,j)×IDF(i,j) (3);

[0067] Based on the SKEP sentiment analysis model, the disagreement index, attention index and sentiment index are constructed; the disagreement index is used to measure the degree of disagreement among investors, and the formula is:

[0068]

[0069] Where B is the number of bullish investors and S is the number of bearish investors. Formula (4) reflects the degree of divergence of opinions among investors on market trends. The calculation formula of the sentiment index is:

[0070] ISI = (V × σ) / S (5);

[0071] In the formula, V represents trading volume, which is usually regarded as an indicator of market activity and uncertainty, S represents investor survey scores, reflecting investor optimism or pessimism, and σ represents the securities market index or price volatility. The calculation formula of the attention index is:

[0072] isi=(V t ×W)+(σ×W)+(else×W) (6);

[0073] Where V t It represents the rate of change of trading volume, and W represents the weight of the corresponding indicator.

[0074] Specifically, natural language processing methods and sentiment analysis models are used to analyze the language structure of the securities market, reconstruct text data, and calculate the TF-IDF value by calculating the word frequency and inverse document frequency. By comprehensively considering the word frequency and inverse document frequency, it can effectively capture important vocabulary in the text, and the calculation speed is fast, which can realize the rapid processing of large data sets.

[0075] At the same time, divergence index, key index and sentiment index are constructed to evaluate the impact of public opinion on the market. The calculation formula of the sentiment index sets multiple pairs of variables, such as market trading volume, stock price volatility and investor survey data. By combining variables, a comprehensive market sentiment index is provided to indirectly reflect investors' attention. The attention index involves the impact of key market activity indicators such as trading volume and price fluctuations. By constructing a sentiment tendency index, the public opinion in the securities market and investor sentiment can be quantified, and the indicator can be incorporated into the design of the subsequent ADRC model to enhance the timeliness of the model's stabilization strategy and the cutting-edge nature of information perception, so as to accurately capture changes in market sentiment, identify risks in a timely manner, and formulate risk management strategies.

[0076] Step S130, constructing an ADRC model based on the target data and the sentiment tendency index.

[0077] Specifically, Figure 2 As shown in the figure, an ADRC (Active Disturbance Rejection Control) model is constructed based on target data and sentiment tendency index. The ADRC model is used to analyze the effectiveness of the stabilization strategy of the virtual simulation model under various conditions, and finally the target stabilization strategy is obtained. The ADRC model can help the securities market to quickly achieve a stable state under severe fluctuations, market surges and market crashes, and can consider the impact of different transition parameters on the stabilization strategy. The AI ​​model is used to dynamically adjust relevant parameters according to the real-time market status to improve the robustness of the system.

[0078] Among them, step S130 includes: constructing a tracking differentiator, an extended state observer and a nonlinear state error feedback control law; wherein the tracking differentiator is used to make the controlled object smoothly transition from the current value to the target value according to the difference between the actual market value and the target value, and the formula is:

[0079]

[0080] In the formula, v1(t) represents the transition process, is the desired stable target, such as the desired market index, and γ is the undetermined parameter that determines the speed of the transition process. The larger the γ, the shorter the time it takes for the transition process to reach the target value. The setting of γ depends on the controller's requirements. The extended state observer is used to obtain securities trend information, and the formula is:

[0081]

[0082] In the formula, z1(t) is the current information of the controlled object, z2(t) and z3(t) reflect the trend information of the controlled object, x1(t) represents the stock market price, α1, α2, δ1, δ2, β 01 ,β 02 ,β 03 , b0 is the ESO unknown parameter, e(t) represents the error of the controlled system, and in order to describe the system reliability, the fal function is set:

[0083]

[0084] In the formula, α and δ are parameters related to the fal function, α is a constant between 0 and 1, and δ is a constant that affects the filtering efficiency and is used to adjust the tracking speed and filtering effect; the nonlinear state error feedback control law formula is:

[0085]

[0086] Where u(t) is the securities market stabilization strategy generated by ADRC, u(t)>0 means buying securities, otherwise it means selling securities, β1, β2 are parameters to be selected, e1(t) represents the current error, and e2(t) represents the error change rate.

[0087] Specifically, Figure 3As shown in the figure, the ADRC model is mainly composed of a tracking differentiator (TD), an extended state observer (ESO) and a nonlinear error feedback control law (NLSEF). Among them, TD solves the control response speed, ESO is used to obtain the trend information of the controlled object, and NLSEF is responsible for eliminating errors and external environmental interference based on the estimated information to obtain a satisfactory control strategy. An ADRC model is constructed that can adapt to the changing characteristics of complex factors in the securities market, can depict market dynamic behavior and effectively respond to various internal and external environmental disturbances.

[0088] By constructing TD, the controlled object can smoothly transition from the current value to the target value according to the difference between the actual market performance and the target value, avoiding market instability caused by the adjustment transition; by constructing ESO to obtain securities trend information, where the future trend of securities can be characterized by variables of the first-order or second-order derivative of time to estimate the output and change speed of the controlled object; NLSEF eliminates errors through estimated information and obtains a satisfactory control strategy. In economic problems, some economic variables are non-negative numbers, and the control variables can be corrected to positive numbers.

[0089] Step S140, according to the target data and the sentiment index, based on the Duffing-Holms model, simulate the stock market under various conditions and construct corresponding virtual simulation models respectively.

[0090] Specifically, in order to simulate the market trading environment, the Duffing-Holms model is used to simulate the random walk state and chaotic state of the securities market according to the target data and sentiment tendency index, and the influence of the control strategy of external interference is taken into account, and the corresponding virtual simulation models are constructed respectively, which is helpful to analyze the effectiveness of the stabilization strategy under different securities market environments, thereby reducing the risk of securities market volatility, helping to maintain the securities market trading order and ensure the stability of the securities market.

[0091] Wherein, step S140 includes: according to the target data, obtaining the Duffing-Holms model, which is:

[0092]

[0093] In the formula, x1(t) represents the stock market price, is the speed of change of the securities market, represents the acceleration of market changes; c represents the first policy parameter, reflecting the intensity of financial system reform and the changes in the exchange rate system; d represents the second policy parameter, reflecting the ability to prevent financial risks, which is measured by foreign exchange reserves, foreign debt and capital structure; z is the speculative disturbance parameter, that is, the disturbance parameter of the impact of large-scale investment or large-scale withdrawal of international investors in the securities market on the market; ω represents the self-regulation frequency of the securities market; a, b are the system rigidity coefficients, that is;

[0094] Based on the theory of chaos model, when a, b, c, d, z, w meet certain conditions, the financial market will change from random walk state to chaotic market crisis state. The Duffing-Holms model of the transition from a random walk state to a chaotic market crisis state is:

[0095]

[0096] The Duffing-Holms model under the influence of external factors is:

[0097]

[0098] In the formula, J represents the jump process amplitude during the violent fluctuation of securities market prices, Q(t) represents the Poisson distribution jump process; σ represents the volatility of securities market prices, and B(t) represents the irregular Brownian motion process; the Duffing-Holms model under the stable strategy is:

[0099]

[0100] In the formula, u(t) represents a stable strategy, u(t)>0 represents the execution of a buy strategy in the market, and u(t)<0 represents the execution of a sell strategy in the market.

[0101] Specifically, after obtaining the target data of the securities market, the Duffing-Holms model is used to simulate the random walk state, chaotic state and collapse state of the coordinated bear market of the securities market according to the target data, so as to analyze the effectiveness of the stabilization strategy under different states.

[0102] Step S150, using the ADRC model to analyze the stabilization strategies of the virtual simulation model under various states to obtain a target stabilization strategy.

[0103] Specifically, according to the virtual simulation models under various states, ADRC is used to analyze the effectiveness of different consolation strategies, and the stabilization strategy with the best effectiveness is selected as the target stabilization strategy, so as to achieve rapid acquisition of stabilization strategies to help the securities market quickly achieve a stable state under severe fluctuations, market surges and market crashes.

[0104] Among them, step S150 includes: obtaining simulation parameters of the ADRC model and bringing them into the ADRC model, the simulation parameters are used to perform initial control on the controlled object; based on the virtual simulation model under various states, using the ADRC model after bringing in the simulation parameters, simulating the smooth strategy, and judging the effectiveness of the smooth strategy, and obtaining the target smooth strategy according to the judgment result.

[0105] Specifically, since the ADRC model has strong robustness and adaptability to nonlinear and strongly uncertain control objects, a set of simulation parameters is first found to perform preliminary control on the controlled object. The relevant parameters are shown in Table 1.

[0106] Table 1 ADRC simulation coefficients

[0107]

[0108]

[0109] In one embodiment, the trend and speed of stock price changes in different stock market environments are simulated.

[0110] According to chaos theory, when a, b, c, d, z, and ω meet certain conditions, the financial market will move from a random walk state to a chaotic market crisis state. When a=-1, b=4, c=2, d=0.077, z=0.044, ω=1.1 are selected, the market is in a state of superposition of chaotic crisis and violent fluctuations. The change of stock market price x1(t) over time is as follows: Figure 4 As shown in (a) of the figure, and its relationship with the changing speed x2(t) is as follows Figure 4 As shown in (b) in .

[0111] If we consider the impact or shock of the external uncertain environment of the securities market, we take σ=10, J=30, and assume that the jump frequency follows the Poisson distribution with intensity parameter λ=0.5, the securities market price x1(t) is as follows Figure 5 As shown in (a) in the figure, the speed of change x2(t) is as follows Figure 5 As shown in (b) of Figure 1, it can be seen that if the impact or shock of the external uncertain environment is taken into account, the securities market system will be in a more unstable chaotic state and the market will fluctuate violently.

[0112] In this embodiment, crawler technology is used to obtain multi-source heterogeneous securities market data, and the securities market data is processed through a data warehouse to obtain target data and store it in the data warehouse, so as to achieve comprehensive and accurate data acquisition, reduce data processing costs and improve collection efficiency. The target data is analyzed through a natural language processing method and a SKEP sentiment analysis model to construct a sentiment index of the securities market, thereby quantifying the securities market public opinion and investor sentiment index, so as to enhance the timeliness and information perception frontier of the stabilization strategy, and accurately capture changes in market sentiment, so as to identify risks in a timely manner; an ADRC model is constructed based on the target data and the sentiment index, and based on the target data and the sentiment index, the securities trading market under various conditions is simulated based on the Duffing-Holms model, and corresponding virtual simulation models are respectively constructed to simulate the market trading environment under various conditions, so as to analyze the effectiveness of the stabilization strategy under various conditions; the ADRC model is used to analyze the stabilization strategy of the virtual simulation model under various conditions to obtain the target stabilization strategy, and the effectiveness of the stabilization strategy is evaluated by analyzing the trend of the securities market, so as to obtain an effective stabilization strategy and improve work efficiency.

[0113] In one embodiment, a comparative analysis may be performed on the stabilization strategies formulated for different market trends to evaluate the effectiveness of the strategy of the ADRC model.

[0114] The effect of stabilization strategy during severe market fluctuations is as follows Figure 6 As shown, (a) represents the change of the changing speed in the stage before the smooth strategy is launched and in the stage after the smooth strategy is launched, (b) represents the changing trend of the securities market price in the stage before the smooth strategy is launched and under the control of the smooth strategy transition process v1(t), (c) represents the change of the changing speed under the control of no smooth strategy and smooth strategy, and (d) represents the case of smooth strategy u(t) and no smooth strategy.

[0115] How effective is the stabilization strategy when the market surges? Figure 7 As shown, (a) represents the change of the change speed in the stage before and after the smooth strategy is started, (b) represents the change trend of the securities market price without a smooth strategy and under the control of the smooth strategy transition process v1(t), (c) represents the change of the change speed without a smooth strategy and under the control of the smooth strategy, and (d) represents the situation of the smooth strategy u(t) and without a smooth strategy.

[0116] How effective is the stabilization strategy during a market crash? Figure 8As shown, (a) represents the change of the changing speed in the stage of not starting the smooth strategy and starting the smooth strategy, (b) represents the changing trend of the securities market price under the control of the transition process v1(t) without the smooth strategy and the smooth strategy, (c) represents the change of the changing speed under the control of the smooth strategy and without the smooth strategy, and (d) represents the situation of the smooth strategy u(t) and without the smooth strategy.

[0117] After introducing the ADRC stabilization strategy through the above three different securities market environments, the market showed a stable fluctuation state after a short period of large fluctuations, indicating that the ADRC model can effectively control the volatility of securities in different market environments.

[0118] In one embodiment, the selection of transition parameters of the ADRC model may directly affect the control accuracy of the ADRC model and the dynamic response performance of the system. Therefore, this embodiment can select the optimal parameters according to the specific requirements of the system and the characteristics of the securities market, combined with the AI ​​model, to achieve the best control effect. The effects of the stabilization strategy with different transition parameters are as follows: Fig. 9 As shown in Figure 1, through three different sets of transition parameters (c), (f) and (i), the price change trend ((a), (d) and (j)) and the change speed trend ((b), (e) and (h)) of the securities market are obtained. In (c), (f) and (i), the transition parameter gradually increases. It can be seen that when the transition stage parameter is larger, the stabilization speed is faster, but the market volatility is larger in the initial stage of stabilization.

[0119] In the process of market stabilization, stabilization costs are bound to be generated. The greater the real-time trading volume, the higher the real-time trading volume. If the adjustment time is too long, the market may not be able to achieve the expected goals in the short term. If the adjustment is too fast, it may significantly increase the real-time costs. This embodiment specifically assumes that the stock trading stabilization cost function is Satisfying the assumption of increasing marginal costs, by drawing the distribution of stable costs and strategy risks, strategy selection and optimization can be achieved accordingly, such as Fig.10 As shown, (a) represents the securities market stabilization cost without disturbance terms, and (b) represents the securities market stabilization cost with disturbance terms.

[0120] In the ADRC model, the selection of transition parameters may directly affect the control accuracy of the ADRC model and the dynamic response performance of the system. Therefore, the present invention selects the optimal parameters according to the specific requirements of the system and the characteristics of the securities market, combines the AI ​​model, achieves the best control effect, enhances the robustness of the ADRC model, optimizes the stabilization strategy, and achieves a balance between risk and return.

[0121] In summary, the present invention can comprehensively consider multiple factors such as the current market situation, stabilization goals and stabilization costs, and obtain a stabilization strategy that is effective in stabilizing the market while taking into account economic costs. Through virtual simulation of the market environment and actual market data testing, strategies for stabilizing the market under different market environments are obtained.

[0122] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0123] Obviously, those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a computer storage medium (ROM / RAM, magnetic disk, optical disk) and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than that here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Therefore, the present invention is not limited to any specific combination of hardware and software.

[0124] The above contents are further detailed descriptions of the present invention in combination with specific implementation methods, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the scope of protection of the present invention.

Claims

1. A method for obtaining a stable strategy for the securities market based on active disturbance rejection control, characterized in that: The following steps are involved: Using crawler technology to obtain multi-source heterogeneous securities market data, and processing the securities market data through a data warehouse to obtain target data and store it in the data warehouse; By using natural language processing methods and SKEP sentiment analysis models, the target data is analyzed to construct a sentiment index of the securities market; Constructing an ADRC model based on the target data and the sentiment tendency index; According to the target data and sentiment index, the stock market under various conditions is simulated based on the Duffing-Holms model, and corresponding virtual simulation models are constructed respectively; The ADRC model is used to analyze the stabilization strategies of the virtual simulation model under various states to obtain a target stabilization strategy.

2. The method for obtaining a stable strategy for the securities market based on active disturbance rejection control according to claim 1, characterized in that: The method of using crawler technology to obtain multi-source heterogeneous securities market data and processing the securities market data through a data warehouse to obtain target data and store it in the data warehouse includes: Obtain multi-source heterogeneous securities market data through public databases, actual securities market transaction data, and crawler technology, and store it in a data warehouse; The securities market data is synchronized, quality tested, cleaned and normalized, summarized, converted and applied through the data warehouse to obtain target data, which is then stored in the data warehouse.

3. The method for obtaining a stable strategy for the securities market based on active disturbance rejection control according to claim 1, characterized in that: The target data is analyzed by using a natural language processing method and a SKEP sentiment analysis model to construct a sentiment index of the securities market, including: The word frequency of securities market public opinion in the target data is calculated by natural language processing method, and the formula is: TF(i,j)=n(i,j) / ∑K n (k,j) (1); In the formula, n(i,j) represents the number of times word t appears in document j, ∑K n (k,j) represents the total number of all feature words in document j; The inverse document frequency of the securities market public opinion in the target data is calculated using the formula: IDF(i,j)=log(|D| / (1+D_t)) (2); In the formula, |D| represents the total number of documents in the corpus, and D_t represents the number of documents containing word t; The TF-IDF value is calculated based on the term frequency and inverse document frequency, and the formula is: TF-IDF(i,j)=TF(i,j)×IDF(i,j) (3); Based on the SKEP sentiment analysis model, the disagreement index, attention index and emotion index are constructed; The divergence index is used to measure the degree of divergence of opinions among investors, and the formula is: In the formula, B is the number of bullish people, and S is the number of bearish people; The calculation formula of the sentiment index is: ISI = (V × σ) / S (5); In the formula, V represents trading volume, S represents investor survey scores, reflecting investor optimism or pessimism, and σ represents volatility; The calculation formula of the attention index is: heat=(V t ×W)+(σ×W)+(else×W) (6); Where V t It represents the rate of change of trading volume, and W represents the weight of the corresponding indicator.

4. The method for obtaining a stable strategy for the securities market based on active disturbance rejection control according to claim 3 is characterized in that: The step of constructing the ADRC model based on the target data and the sentiment index includes: Construct tracking differentiator, extended state observer and nonlinear state error feedback control law; The tracking differentiator is used to make the controlled object smoothly transition from the current value to the target value according to the difference between the actual market value and the target value. The formula is: In the formula, v1(t) represents the transition process, is the desired stable target, γ is the undetermined parameter that determines the speed of the transition process. The larger the γ, the shorter the time it takes for the transition process to reach the target value. The extended state observer is used to obtain securities trend information, and the formula is: In the formula, z1(t) is the current information of the controlled object, z2(t) and z3(t) reflect the trend information of the controlled object, x1(t) represents the stock market price, α1, α2, δ1, δ2, β 01 ,β 02 ,β 03 , b0 is the ESO unknown parameter, e(t) represents the error of the controlled system, and the fal function is: Wherein, α and δ are parameters related to the fal function; the nonlinear state error feedback control law formula is: Where u(t) is the securities market stabilization strategy generated by ADRC, u(t)>0 indicates a buy strategy, otherwise it indicates a sell strategy, β1, β2 are parameters to be selected, e1(t) indicates the current error, and e2(t) indicates the error change rate.

5. The method for obtaining a stable strategy for the securities market based on active disturbance rejection control according to claim 4, characterized in that: According to the target data and the sentiment index, the stock market under various conditions is simulated based on the Duffing-Holms model, and corresponding virtual simulation models are constructed respectively, including: According to the target data and sentiment index, the Duffing-Holms model is obtained as follows: In the formula, x1(t) represents the stock market price, is the speed of change of the securities market, represents the acceleration of market changes; c represents the first policy parameter; d represents the second policy parameter; z is the speculative disturbance parameter; ω represents the self-regulation frequency of the securities market; a and b are the system rigidity coefficients; make The Duffing-Holms model of the transition from random walk state to chaotic market crisis state is: The Duffing-Holms model under the influence of external factors is: In the formula, J represents the jump process amplitude during the violent fluctuation of securities market prices, Q(t) represents the Poisson distribution jump process; σ represents the volatility of securities market prices, and B(t) represents the irregular Brownian motion process; The Duffing-Holms model under the stationary strategy is: In the formula, u(t) represents a stable strategy, u(t)>0 represents the execution of a buy strategy in the market, and u(t)<0 represents the execution of a sell strategy in the market.

6. The method for obtaining a stable strategy for the securities market based on active disturbance rejection control according to claim 5, characterized in that: The ADRC model is used to analyze the stabilization strategy of the virtual simulation model under various states to obtain a target stabilization strategy, including: Acquiring simulation parameters of an ADRC model and bringing them into the ADRC model, wherein the simulation parameters are used to perform preliminary control on the controlled object; Based on virtual simulation models under various states, the ADRC model with simulation parameters is used to simulate the stabilization strategy, and the effectiveness of the stabilization strategy is judged. The target stabilization strategy is obtained according to the judgment results.

Images