Market collusion frying risk monitoring simulation system, method and related device

Abstract

The invention discloses a market collusion frying risk monitoring simulation system and method and a related device. The system comprises a system parameter setting module used for setting basic parameters of a market environment; the agent parameter setting module is used for configuring key parameters of various agents in the market; the market initialization and calibration module is used for generating an initial state and intelligent agent distribution of a simulation market based on the basic parameters and the key parameters; the agent behavior simulation module is used for simulating transaction behaviors of various agents in the market; the collusion stir-frying strategy implementation module is used for simulating the stir-frying behavior of a stir-frying person; and the stir-frying identification module is used for adjusting key parameters of a collusion stir-writer and an associated account of the collusion stir-writer to obtain stir-frying behavior sample data of different stir-frying intensities, and identifying stir-frying scenes of different stir-frying intensities through a statistical distribution analysis method or a machine learning method. According to the method and the device, the recognition accuracy of the stir-frying behavior of multi-account cooperative operation is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a market collusion speculation risk monitoring simulation system, method and related devices. Background Art

[0002] Market speculation has become a prominent problem that seriously affects market fairness and transparency. In particular, in the stock market, collusive speculators represented by private equity funds use complex multi-account coordination strategies and hidden trading operations to manipulate market prices and trading volumes in pursuit of short-term profit maximization. This kind of speculation not only harms the interests of other investors and disrupts the normal market order, but also poses a systemic threat to the overall stability of the financial market.

[0003] Traditional market speculation detection methods mainly rely on the identification of abnormal trading features, and judge whether there is speculation by matching preset trading patterns. Such methods are usually based on the monitoring of abnormal price fluctuations and surges in trading volume. However, as the concealment and complexity of collusion speculation strategies continue to increase, these static methods are difficult to timely and accurately identify their complex operating patterns. Collusion speculators often use multiple accounts to disperse transactions and further conceal their true intentions through synchronized market operations, thereby bypassing traditional rule detection. The main drawback of this static analysis is that it cannot capture the interaction dynamics between market participants and lacks a deep understanding of the market price formation mechanism, which leads to easy omissions in complex market environments.

[0004] In recent years, with the rapid development of machine learning and big data technology, market speculation detection has entered a data-driven intelligent stage. For example, supervised learning models are widely used in speculation detection tasks. Classification models are trained with a large amount of labeled speculation behavior data to predict and detect abnormal market trading behaviors. However, such methods are highly dependent on high-quality data. In practice, it is extremely difficult to obtain accurately labeled speculation behavior data, and it is difficult to effectively identify speculation strategies involving multi-account collaborative operations, and it is impossible to capture the real intentions of speculators concealed by decentralized transactions and synchronous operations. Summary of the invention

[0005] The present application provides a market collusion speculation risk monitoring simulation system, method and related devices, which are used to improve the technical problem that the existing technology cannot effectively identify speculation behaviors involving collaborative operations of multiple accounts, resulting in easy omissions in complex market environments.

[0006] In view of this, the first aspect of the present application provides a market collusion speculation risk monitoring simulation system, comprising:

[0007] A system parameter setting module, used to set basic parameters of the market environment, including initial market price, price volatility, time step and transaction fee;

[0008] An agent parameter setting module, used to configure key parameters of various agents in the market, including retail investors, colluding speculators, and associated accounts of colluding speculators;

[0009] A market initialization and calibration module, used to generate the initial state and agent distribution of the simulated market based on the basic parameters and the key parameters;

[0010] Agent behavior simulation module, used to simulate the trading behavior of various agents in the market;

[0011] The module for implementing the collusion hype strategy is used to simulate the hype behavior of the hype man;

[0012] The hype identification module is used to adjust the key parameters of the colluding hype person and the associated account of the colluding hype person to obtain hype behavior sample data of different hype intensities, and identify hype scenarios of different hype intensities through statistical distribution analysis methods or machine learning methods.

[0013] Optionally, the key parameters of the retail investor include maximum investment period, technical belief weight, risk aversion coefficient, fundamental volatility, historical price volatility, initial wealth, and follow-up probability adjustment factor;

[0014] The key parameters of the associated accounts of the collusive speculators include the manipulation signal adjustment factor, initial wealth, order size adjustment factor, and technical belief weight;

[0015] The key parameters of the collusive speculators include technical belief weight, speculation strategy selection factor, risk aversion coefficient, manipulation signal adjustment factor, regulatory risk adjustment factor and initial wealth.

[0016] Optionally, the market initialization and calibration module is specifically used to:

[0017] Generate an initial price sequence according to the initial market price and the price volatility to establish an initial price environment of the market;

[0018] Allocate funds to various intelligent entities according to the initial wealth, and determine the initial trading capacity of each intelligent entity;

[0019] Adjust market parameters, number of agents, and behavioral rules through iterative simulation until the formatting characteristics of the artificial stock market match real market data.

[0020] Optionally, the agent behavior simulation module includes a retail investor behavior simulation submodule, and the retail investor behavior simulation submodule is used to:

[0021] Determining the investment period of the retail investor based on the maximum investment period of the retail investor and the technical belief weight;

[0022] Combining the moving average of historical prices and the recent price trend, and introducing random belief errors, determine the expected price of the retail investors;

[0023] Calculate the risk level faced by the retail investors by combining the fundamental volatility, historical price volatility and the trend-following probability adjustment factor of the retail investors;

[0024] Determining the absolute risk aversion coefficient of the retail investor based on the technical belief weight and risk aversion coefficient of the retail investor;

[0025] Calculating the order submission price of the retail investor based on the retail investor's expected price, risk level and absolute risk aversion coefficient;

[0026] Determining the optimal wealth investment ratio of the retail investors based on the expected spot yield, risk level faced and absolute risk aversion coefficient of the retail investors;

[0027] The order size of the retail investor is calculated based on the retail investor's optimal wealth investment ratio, current wealth, order submission price and current position quantity.

[0028] Optionally, the agent behavior simulation module includes a submodule for simulating the behavior of an associated account of a colluding speculator, and the submodule for simulating the behavior of an associated account of a colluding speculator is used to:

[0029] Simulate the trading behavior of the retail investors, and add the speculators' adjustment items to the expected price of the retail investors based on the manipulation signal adjustment factor to obtain the expected price of the associated accounts of the colluding speculators; and add the additional trading volume specified by the speculators to the order size of the retail investors based on the order size adjustment factor to obtain the order size of the associated accounts of the colluding speculators.

[0030] Optionally, the agent behavior simulation module includes a collusion speculator behavior simulation submodule, and the collusion speculator behavior simulation submodule is used to:

[0031] Determine the absolute risk aversion coefficient of the colluding speculators according to the technical belief weight and risk aversion coefficient of the colluding speculators;

[0032] Calculate the order submission price of the colluding speculators according to the expected price, risk level faced and absolute risk aversion coefficient of the colluding speculators;

[0033] Determine the optimal wealth investment ratio of the colluding speculators according to the expected spot rate of return, the risk level faced and the absolute risk aversion coefficient of the colluding speculators;

[0034] The order size of the colluding speculator is calculated based on the colluding speculator's optimal wealth investment ratio, current wealth, order submission price and current position quantity.

[0035] Optionally, the system further includes: a regulatory measure application module, which is used to simulate different regulatory strategies to adjust regulatory parameters so as to intervene in the market and curb speculation;

[0036] Specifically, it is used to increase regulatory risk adjustment factors and penalties, so that speculators consider the detected risks when making decisions and adjust the corresponding expected returns;

[0037] Set a transaction volume cap. When the transaction volume exceeds the threshold, the simulation reporting mechanism generates a report;

[0038] Reduce information disclosure delays and make position information public;

[0039] Adjust the risk appetite and follow-up probability of the retail investors;

[0040] Set price fluctuation limits and introduce circuit breaker mechanisms.

[0041] Optionally, the adjustment process of the regulatory parameters includes:

[0042] Taking the regulatory parameters as variables, randomly initializing and generating a plurality of regulatory parameter combinations within a preset range;

[0043] Conduct market simulation based on each of the regulatory parameter combinations, calculate fitness values ​​based on the simulation results, and select several excellent regulatory parameter combinations based on the fitness values; the fitness values ​​include transaction anomaly, price volatility, or speculator profit level;

[0044] Genetic operations are performed on the excellent regulatory parameter combinations to generate new regulatory parameter combinations, and the iterations are repeated until a stop condition is met, and the optimal regulatory parameter combination is output.

[0045] Optionally, the system also includes a market feedback and result analysis module, which is used to collect and analyze simulation results and evaluate the effects of speculation and regulatory measures;

[0046] It is specifically used to count the number of times speculators speculate under different regulatory conditions;

[0047] Assess price volatility and market stability;

[0048] Analyze the average returns and risk tolerance of retail investors under different conditions;

[0049] Assess the impact of regulatory measures on speculators’ profit expectations and risks;

[0050] Determine whether regulatory measures have improved the market environment.

[0051] Optionally, the system further includes a regulatory agency agent, which is used to dynamically adjust regulatory strategies based on market data and the behavior of each agent.

[0052] Optionally, the hype identification module is specifically used to:

[0053] Adjust the key parameters of the colluding speculators and the associated accounts of the colluding speculators to obtain sample data of speculation behaviors with different speculation intensities, compare the distribution differences of speculation scenarios with different speculation intensities on market indicators through distribution difference measurement, and use statistical distance measurement to determine the degree of deviation caused by the speculation behaviors to the overall market distribution, so as to identify speculation scenarios with different speculation intensities;

[0054] Alternatively, the market indicators of the sample data of the hype behavior at each hype intensity are mapped to feature vectors, and then the hype scenarios of different hype intensities are identified through a clustering model or a classification model.

[0055] Optionally, the speculation identification module is further used to obtain account data samples with different speculation characteristics by adjusting the order size, order time distribution and position change rules of each account in the simulated market;

[0056] Extracting feature representations of each account from the account data sample through a graph neural network, classifying the feature representations of each account through a classifier or clustering the feature representations of each account through a clustering model, and obtaining collusion speculation accounts;

[0057] Or, quantitatively describe the account data samples in different feature dimensions to construct an account profile and extract feature indicators for distinguishing normal transactions from suspicious transactions. The feature dimensions include transaction frequency, buying and selling sequence, and correlation with market price fluctuations;

[0058] Perform graph association analysis, correlation measurement statistics or machine learning model classification on the characteristic indicators to identify abnormal linkage relationships between accounts and obtain collusion speculation accounts.

[0059] Optionally, the training process of the graph neural network includes:

[0060] Take the accounts as nodes, construct edges based on the back-to-back transactions and fund flows between accounts, and attach the position, order frequency, transaction amount and / or correlation to each node and its edge to obtain a graph structure;

[0061] The association representation between accounts is learned based on node and edge features through graph convolutional networks.

[0062] Optionally, the system further includes: a dynamic risk assessment module, for:

[0063] A set of nodes reflecting speculation risk is defined in the Bayesian network. The nodes include market volatility, historical price anomaly, transaction correlation between collusion accounts, regulatory intensity and / or information disclosure delay. The conditional dependency relationship between the nodes is represented by directed edges.

[0064] For each node, a conditional probability table is constructed based on the corresponding parent node;

[0065] When the external environment or market status changes, updated data is obtained from real-time monitoring or simulation results, and the posterior probability of each node is updated based on Bayesian theorem;

[0066] The risk level of collusion speculation in the current market scenario is determined based on the posterior probability of each node.

[0067] The second aspect of the present application provides a market collusion speculation risk monitoring simulation method, comprising:

[0068] Setting the basic parameters of the market environment and configuring the key parameters of various intelligent agents in the market, the basic parameters include the initial market price, price volatility, time step and transaction fees, and the intelligent agents include retail investors, colluding speculators and the associated accounts of colluding speculators;

[0069] Generate the initial state and agent distribution of the simulated market based on the basic parameters and the key parameters;

[0070] Simulate the trading behaviors of various intelligent agents in the market;

[0071] Simulate the hype behavior of speculators;

[0072] The key parameters of the colluding speculator and the associated account of the colluding speculator are adjusted to obtain sample data of speculation behavior with different speculation intensities, and the speculation scenarios with different speculation intensities are identified through statistical distribution analysis methods or machine learning methods.

[0073] A third aspect of the present application provides an electronic device, the device comprising a processor and a memory;

[0074] The memory is used to store program codes and transmit the program codes to the processor;

[0075] The processor is used to execute the market collusion speculation risk monitoring simulation method described in the second aspect according to the instructions in the program code.

[0076] The fourth aspect of the present application provides a computer-readable storage medium, which is used to store program code. When the program code is executed by a processor, it implements the market collusion speculation risk monitoring simulation method described in the second aspect.

[0077] It can be seen from the above technical solutions that this application has the following advantages:

[0078] The market collusion speculation risk monitoring simulation system provided by the present application, by constructing a multi-agent model including retail investors, collusion speculators and their associated accounts, deeply simulates the decentralized trading and synchronous operation strategies of collusion speculators, accurately captures their hidden true intentions, and helps to improve the detection accuracy of collusion speculation behavior; by adjusting the key parameters of collusion speculators and their associated accounts for market simulation, speculative behavior sample data of different speculation intensities are generated, which makes up for the problem of low accuracy of speculative behavior detection caused by the difficulty in obtaining speculative behavior data in the prior art. In addition, the present application generates speculative behavior sample data of different intensities, and combines statistical distribution analysis or machine learning methods to extract the significant features of speculative behavior, thereby effectively improving the ability to identify multi-account collusion operations. Compared with the traditional method that only relies on a single abnormal threshold or rule matching, the present application can achieve more flexible and precise identification of collusion speculation strategies of different intensities and types, which helps to timely discover potential coordinated manipulation risks in real regulatory scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0080] Figure 1 A structural schematic diagram of a market collusion speculation risk monitoring simulation system provided in an embodiment of the present application;

[0081] Figure 2 Another structural schematic diagram of a market collusion speculation risk monitoring simulation system provided by an embodiment of the present application;

[0082] Figure 3 A flow chart of a market collusion speculation risk monitoring simulation method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0083] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0084] For easier understanding, please refer to Figure 1 The embodiment of the present application provides a market collusion speculation risk monitoring simulation system, including:

[0085] System parameter setting module, used to set the basic parameters of the market environment, including initial market price, price volatility, time step and transaction fee;

[0086] The agent parameter setting module is used to configure the key parameters of various agents in the market. The agents include retail investors, collusion speculators, and the associated accounts of collusion speculators. The key parameters include at least technical belief weights and initial wealth.

[0087] Market initialization and calibration module, used to generate the initial state and agent distribution of the simulated market based on basic parameters and key parameters;

[0088] Agent behavior simulation module, used to simulate the trading behavior of various agents in the market;

[0089] The module for implementing the collusion hype strategy is used to simulate the hype behavior of the hype man;

[0090] The hype identification module is used to adjust the key parameters of the colluding speculators and the associated accounts of the colluding speculators to obtain sample data of hype behaviors with different hype intensities, and identify hype scenarios with different hype intensities through statistical distribution analysis methods or machine learning methods.

[0091] In the embodiment of the present application, the simulation system first performs basic configuration through the system parameter setting module and the agent parameter setting module, and then generates the initial market state in the market initialization and calibration module. Next, the agent behavior simulation module simulates the trading behavior of market participants, and the collusion speculation strategy implementation module simulates the speculation behavior of speculators. Subsequently, the regulatory measures application module intervenes in the market according to the set regulatory strategy. Finally, the market feedback and result analysis module evaluates and analyzes the simulation results.

[0092] The system parameter setting module is used to set the basic parameters of the market environment to ensure that the simulated market can reflect the characteristics of the real market. The basic parameters include:

[0093] 1) Initial market price :Used to determine the initial price level of the market, which is the starting point of the simulated market;

[0094] 2) Price volatility :Determines the basic volatility of market prices and affects the random fluctuation range of prices;

[0095] 3) Time step T: sets the total time period of the simulation and determines the length of the simulation process;

[0096] 4) Transaction cost c: It represents the proportion of fees generated during the transaction process, reflecting the impact of transaction costs on the market.

[0097] The agent parameter setting module is used to configure the key parameters of various agents in the market, including specific parameters of retail investors, collusion speculators, and the associated accounts of collusion speculators. The setting of these parameters has an important impact on the simulation results and needs to be reasonably configured according to the actual market situation.

[0098] Retail investors are major players in the market and their behavior has a significant impact on market dynamics. Key parameters of retail investors include:

[0099] 1) Maximum investment period : The maximum investment period of retail investor i, which determines the length of time he holds a position and reflects the short-term or long-term nature of his investment strategy;

[0100] 2) Technical Belief Weight : The degree of reliance of retail investor i on technical information, which affects the basis of his trading decision, with a value range of [0,1];

[0101] 3) Risk aversion coefficient : The risk aversion of retail investor i, which affects the risk level of its investment portfolio, with a typical value in the range [0.5,5];

[0102] 4) Fundamental volatility : It is used to measure the risk metric based on fundamental signals, which affects the risk assessment of retail investors i;

[0103] 5) Historical price volatility : Risks estimated based on historical price changes, affecting retail investors’ expectations of market volatility;

[0104] 6) Initial Wealth : Set the initial capital amount of retail investor i to affect his trading ability;

[0105] 7) Follow-up Probability Adjustment Factor : Determines the probability of whether retail investor i will follow the trend and affects the tendency of his trading behavior.

[0106] The actions of the associated accounts of the colluding speculators are carried out under the instructions of the speculators. The key parameters of the associated accounts of the colluding speculators include:

[0107] 1) Control signal adjustment factor : The influence of the operation instructions transmitted by the speculator through the associated account i on the behavior of the proxy retail investors, with a value range of [0,1];

[0108] 2) Initial Wealth : The initial amount of funds in account i can be the same as that of retail investors or appropriately increased;

[0109] 3) Order size adjustment factor : Affects the order size of the associated account i, which has the effect of enlarging or reducing the transaction volume;

[0110] 4) Technical Belief Weight : The degree of reliance of associated account i on technical analysis can be adjusted according to the instructions of the speculator.

[0111] Collusive speculators are the leaders of market speculation, and their parameter settings are crucial to the simulation results. The key parameters of collusive speculators include:

[0112] 1) Technical Belief Weight :The degree to which the speculator CM relies on market signals affects the formulation of its speculation strategy;

[0113] 2) Hype strategy selection factor: determines whether the speculator adopts single-point hype (0) or multi-point hype (1), affecting the complexity and concealment of the hype behavior;

[0114] 3) Risk aversion coefficient : The risk aversion of the speculator CM affects the risk-taking level during the speculation process and usually takes a lower value;

[0115] 4) Control signal adjustment factor :The intensity of the instructions transmitted by the speculator CM to the associated accounts affects the trading behavior of the associated accounts;

[0116] 5) Regulatory Risk Adjustment Factor : Considering the probability of risk being detected by regulatory agencies, it affects the expected returns and strategy choices of speculators;

[0117] 6) Initial Wealth :Set the initial capital amount of the speculator CM, which is usually larger to reflect its dominant position in the market.

[0118] The market initialization and calibration module is used to generate the initial state and agent distribution of the simulated market to ensure that the simulated market has statistical characteristics consistent with the real market. The market initialization and calibration module is specifically used to generate the initial state and agent distribution of the simulated market according to the set initial market price. and price volatility Generate the initial price sequence and establish the initial price environment of the market; allocate funds to various agents according to the initial wealth set in the agent parameter setting module, and determine the initial trading capacity of each agent; adjust market parameters, the number of agents and behavioral rules, and continuously iterate the simulation until the formatting characteristics of the artificial stock market (such as sharp peaks and thick tails) match the real market data, ensuring the accuracy and reliability of subsequent simulations.

[0119] The agent behavior simulation module is the core part of the simulation system, which is used to simulate the trading behaviors of various agents in the market. Different types of agents have their own trading strategies and decision-making processes, and their behaviors have different impacts on market dynamics.

[0120] The agent behavior simulation module includes a retail investor behavior simulation submodule, which is used to:

[0121] Determine the investment period of retail investors based on their maximum investment period and technical belief weights;

[0122] Combine the moving average of historical prices and the recent price trend, and introduce random belief errors to determine the expected price of retail investors;

[0123] Calculate the risk level faced by retail investors by combining their fundamental volatility, historical price volatility and the adjustment factor for the probability of following the trend;

[0124] Determine the absolute risk aversion coefficient of retail investors based on their technical belief weights and risk aversion coefficients;

[0125] Calculate the price at which retail investors submit their orders based on their expected price, risk level, and absolute risk aversion coefficient;

[0126] Determine the optimal wealth investment ratio for retail investors based on their expected spot yield, risk level, and absolute risk aversion coefficient;

[0127] The order size of retail investors is calculated based on their optimal wealth investment ratio, current wealth, order submission price and current position quantity.

[0128] Specifically, when simulating the behavior of retail investors, the retail investors enter the market randomly according to the Poisson process, simulating the characteristics of their random transactions. Determined by the following formula:

[0129]

[0130] in, is the maximum investment period of retail investor i, is the technical belief weight of retail investor i. The above formula reflects the impact of technical belief on investment period. The stronger the technical belief, the shorter the investment period.

[0131] During the trading process, retail investors determine the trading volume and price based on their expected price and risk assessment. By combining fundamental and technical information, the moving average of historical prices and the recent price trend are taken into account, while introducing random belief errors. , simulating the cognitive bias of retail investors on market information, namely:

[0132]

[0133] in, is the fundamental value, calculated as the moving average of historical prices; is the price at the current time t; is the potential belief error about the fundamental value, which follows a uniform distribution .

[0134] Risk measurement using variance , combined with fundamental volatility and historical price volatility , calculate the risk level faced by retail investor i:

[0135]

[0136] in, is the follow-up probability adjustment factor, and the calculation formula is:

[0137]

[0138] Optimize retail investors’ investment portfolios and determine the optimal wealth investment ratio :

[0139]

[0140] in, is the expected spot rate of return, and the calculation formula is:

[0141]

[0142] Absolute risk aversion coefficient It is expressed as:

[0143]

[0144] in, The risk aversion coefficient of retail investor i set in the agent parameter setting module;

[0145] Order size According to the optimal portfolio of retail investor i and the current number of positions calculate:

[0146]

[0147] Order Submission Price for:

[0148]

[0149] in, , represents a random price disturbance.

[0150] The agent behavior simulation module also includes a submodule for simulating the behavior of the associated accounts of the colluding speculators. The submodule for simulating the behavior of the associated accounts of the colluding speculators is used to:

[0151] Simulate the trading behavior of retail investors, add the speculator's adjustment item to the expected price of retail investors based on the manipulation signal adjustment factor, and get the expected price of the associated account of the colluding speculator; based on the order size adjustment factor, add the additional trading volume specified by the speculator to the order size of retail investors, and get the order size of the associated account of the colluding speculator.

[0152] The associated accounts of the colluding speculators simulate the trading behavior of retail investors under the instructions of the colluding speculators, but their expected prices and order sizes are affected by the colluding speculators. The expected prices are adjusted by the colluding speculators on the basis of the original retail investors. :

[0153]

[0154] The order size is also based on the original retail investors plus the additional trading volume specified by the collusion speculators. :

[0155]

[0156] in, and They respectively represent the collusion of speculators to the associated accounts Adjustments in expected prices and order sizes.

[0157] The intelligent agent behavior simulation module also includes a collusion speculator behavior simulation submodule, which is used to:

[0158] Determine the absolute risk aversion coefficient of the collusive speculators according to their technical belief weights and risk aversion coefficients;

[0159] Calculate the order submission price of the colluding speculators based on their expected price, risk level and absolute risk aversion coefficient;

[0160] Determine the optimal wealth investment ratio of the collusive speculators based on their expected spot yield, risk level and absolute risk aversion coefficient;

[0161] The order size of the colluding speculators is calculated based on their optimal wealth investment ratio, current wealth, order submission price and current position quantity.

[0162] Specifically, collusive speculators decide on their speculation methods and trading behaviors based on market conditions and their own strategies. The calculation of the expected price and order size of collusive speculators is similar to that of retail investors, but takes into account the particularity of speculation strategies and regulatory risks. Absolute risk aversion coefficient It is expressed as:

[0163]

[0164] The order size is calculated as:

[0165]

[0166] in, is the expected rate of return of the colluding speculator CM, is the current holdings of the colluding speculators, The level of risk faced by CMs who collude to speculate.

[0167] The order submission price of the colluding speculator CM for:

[0168]

[0169] in, Represents the price disturbance caused by colluding speculators.

[0170] The collusion speculation strategy implementation module is used to simulate the specific speculation behavior of speculators, including single-point speculation and multi-point speculation. In single-point speculation, collusion speculators independently speculate in the market through their own accounts without using associated accounts. The speculation process includes:

[0171] S1. Accumulation stage: speculators gradually buy at low levels to build positions and avoid attracting market attention;

[0172] S2. Pull-up stage: speculators buy with larger orders and higher prices, pushing up market prices and creating an upward trend;

[0173] S3. Selling stage: speculators gradually sell their positions at high levels to gain profits;

[0174] In multi-point speculation, speculators use their own accounts and multiple associated accounts to coordinate market speculation, increasing the concealment of speculation. The speculation process includes:

[0175] S1. Fund-accumulation stage: speculators and related accounts jointly buy at low levels to diversify their positions and avoid attracting market attention;

[0176] S2. Pull-up phase: Speculators and related accounts buy at different times, prices, and quantities in a coordinated manner to push up prices and create the illusion of an active market;

[0177] S3. Shipping stage: Speculators and related accounts sell their positions in batches to maintain price stability and avoid sudden price drops.

[0178] The speculation identification module adjusts the key parameters of the collusive speculators and their associated accounts (such as speculation strategy selection factor, order size adjustment factor, risk aversion coefficient, etc.), and gradually generates multiple sets of artificial market simulation data from low to high speculation levels. These speculation scenarios with different speculation intensities can reflect the market-level changes (such as price distribution characteristics, trading volume distribution characteristics, position distribution characteristics, etc.) brought about by the collusive speculators implementing different levels of speculation in the market. Specifically, the speculation intensity can be controlled by parameter adjustment to obtain speculation sample data with different speculation intensities; in each simulation, for the collusive speculators and their associated accounts, the key parameters such as the manipulation signal adjustment factor, order size adjustment factor, and speculation strategy selection factor are successively increased or decreased, so that they can change from "almost no intervention in the market" to "frequent and large-scale intervention in the market". Subsequently, based on these multi-level and multi-dimensional market indicators, different intensities of speculation can be identified in the distribution or distance space. For example, the distribution differences of various intensity speculation scenarios such as non-hype / weak speculation scenarios, medium speculation, and strong speculation scenarios can be compared through distribution difference measurement, and statistical distance measurement (such as KL divergence, JS divergence, Wasserstein distance, etc.) can be used to judge the degree of deviation caused by speculation to the overall market distribution; or after mapping the market indicators of each scenario to feature vectors, clustering methods (such as K-means, DBSCAN, etc.) or classification methods (such as SVM, random forest, etc.) can be used to distinguish speculation scenarios of different speculation intensities. It should be noted that the market indicators in the embodiments of the present application refer to a variety of statistical or time series parameters that can quantify the overall dynamic characteristics of the market, which are used to characterize the differences in price and transaction characteristics presented by the market under different speculation intensities. For example, it may include price distribution indicators (such as the mean, variance, skewness, kurtosis and time series autocorrelation coefficient of the logarithmic return distribution), trading volume distribution indicators (such as the concentration of trading volume in different time windows, extreme value ratio and transaction cluster aggregation), volatility indicators (such as intraday or multi-day volatility, variance series estimated by the GARCH model), and account position distribution indicators (such as the concentration of positions after merging multiple accounts, and the skewness or kurtosis of the position market value distribution).

[0179] Specifically, the distribution difference metric in the embodiment of the present application mainly compares and quantitatively evaluates the statistical characteristics presented by hype scenarios of different intensities on various market indicators. First, different hype scenarios can be obtained based on hype sample data of different hype intensities, and the market indicators sampled in each hype scenario can be modeled for probability distribution, for example, by using histogram estimation, kernel density estimation (KDE) or fitting a specific distribution family (such as normal distribution, lognormal distribution, generalized Pareto distribution, etc.) to obtain the distribution function of each hype scenario. Subsequently, for the distribution functions of non-hype or weak hype scenarios and strong hype scenarios on the same market indicators, the degree of difference between their distributions is calculated by statistical distance metrics (such as Kullback-Leibler divergence, Jensen-Shannon divergence or Wasserstein distance, etc.). After obtaining the distribution differences of each hype scenario on the market indicators, the above distance metric value can be used as a quantitative measurement basis for the degree of deviation. The larger the distance between the distributions, the more obvious the difference between the characteristics of the market indicator in the two scenarios, which can better reflect the degree of market disturbance caused by speculation of different intensities. For example, if a strong speculation scenario shows a large distribution difference in price fluctuation characteristics, frequency of extreme trading volume events, etc. compared with the benchmark scenario (non-speculation or weak speculation), it indicates that the speculation in this scenario has a more significant impact on the market.

[0180] After clarifying the distribution deviation between different scenarios, we can set several thresholds or use clustering / classification methods to identify different intensity hype scenarios. Specifically, we can construct a feature vector d based on the multidimensional distribution distance, such as , where each component They respectively represent the distribution differences of market indicators such as price distribution, trading volume distribution, and volatility between the two scenarios; next, an unsupervised clustering algorithm (such as K-means or DBSCAN) can be used to automatically divide different cluster centers in the feature space to form a division of "non-hype / weak hype", "medium hype", and "strong hype" scenarios; supervised classification models (such as SVM, random forest, etc.) can also be used to train the feature vector d to achieve rapid discrimination of the hype intensity of the new scenario. When the distribution difference exceeds a certain threshold, the scenario can be identified as a high-level hype scenario. If the degree of difference is small, it may belong to a non-hype or weak hype scenario. In this way, the system can not only distinguish between collusion hype behaviors of different intensities, but also quantify the specific impact of collusion hype on the market, providing a more fine-grained reference for regulatory decisions.

[0181] Through the above-mentioned method, on the one hand, artificial market simulation is used to generate multi-level speculation behavior data, and on the other hand, statistical distribution analysis or machine learning methods are combined to extract the significant features of speculation behavior, thereby effectively improving the ability to identify multi-account collusion operations; compared with the traditional method that only relies on a single abnormal threshold or rule matching, this embodiment can achieve more flexible and precise identification of collusion speculation strategies of different intensities and types, which is helpful to timely discover potential collaborative manipulation risks in real regulatory scenarios.

[0182] Furthermore, the speculation identification module is also used to obtain account data samples with different speculation characteristics by adjusting the order size, order time distribution and position change rules of each account in the simulated market;

[0183] Extract the feature representation of each account from the account data sample through the graph neural network, classify the feature representation of each account through the classifier or cluster the feature representation of each account through the clustering model to obtain the collusion speculation account;

[0184] Or, quantitatively describe account data samples in different feature dimensions to construct account portraits and extract feature indicators for distinguishing normal transactions from suspicious transactions. Feature dimensions include transaction frequency, buying and selling sequence, and correlation with market price fluctuations;

[0185] Conduct graph association analysis, correlation measurement statistics, or machine learning model classification on feature indicators to identify abnormal linkage relationships between accounts and obtain accounts that collude to speculate.

[0186] In order to more precisely locate and judge the speculation behavior of collusive speculators and their associated accounts at the account level, we can combine the time series characteristics of multi-account transaction data, order patterns, transaction network structure and other information to further conduct account-level analysis on the generated speculation scenarios of different intensities. Specifically, by adjusting the order size, order time distribution and position change rules of each account in the artificial market simulation environment, we can form multiple account data samples with different speculation characteristics.

[0187] In one embodiment, each account can be quantitatively described in different dimensions (such as transaction frequency, buying and selling sequence, correlation with market price fluctuations, etc.) to construct an account portrait and extract characteristic indicators that can distinguish normal transactions from suspicious transactions. Subsequently, based on these characteristic indicators, graph association analysis, statistical correlation measurement or machine learning classification model and other methods can be used to identify abnormal linkage relationships between accounts, thereby discovering possible multi-account collaborative operations. If certain account groups repeatedly conduct large transactions in the same direction in a very short period of time and the transaction returns are highly similar, or maintain a synchronous trading rhythm under different market conditions, it can be considered that the account group is suspected of speculation. In addition, the above results can be further tested for robustness to test the accuracy and robustness of the identification method under different simulation scenarios or disturbance conditions. Through the above-mentioned refined analysis at the account level, hidden signs of collusion between multiple accounts can be captured at a more micro level, the ability to identify complex and hidden speculation behaviors can be improved, and more targeted detection and evidence support can be provided to regulators.

[0188] In another embodiment, collusion speculation behavior can be detected based on graph neural network to identify collusion speculation accounts. First, accounts can be regarded as nodes in the graph, and edges can be established between nodes based on multiple back-to-back transactions, capital flows, and other relationships. Features such as position size, order frequency, transaction amount, and correlation are added to each node and its edge, so as to construct a graph structure that can reflect the real interaction mode between accounts. In the training stage, models such as Graph Convolutional Network (GCN) or Graph Attention Network (GAT) are selected to learn the association representation between accounts based on node and edge features; in addition, based on time series features, transaction data can be divided into several snapshot graphs according to time periods or transaction events, and each snapshot graph can be embedded and fused in time series to capture dynamic evolution characteristics. After the training of the graph neural network is completed, the node feature representation (i.e., the feature representation of the account) output by it is input into the subsequent classifier to identify suspicious accounts or predict whether the account has a tendency to collude in transactions; it is also possible to find abnormally dense subgraphs in the embedding space based on clustering methods and lock in account groups that may participate in collusion speculation, providing real-time and accurate clues for regulators.

[0189] Specifically, in an optional embodiment of the present application, for snapshot graphs formed by multiple accounts at different time periods (or transaction events), graph neural network models (such as GCN or GAT) can be used to extract graph embedding features for nodes and edge information in each snapshot graph, and obtain node vector representation or full graph vector representation at each moment. Subsequently, the embedded features obtained at each moment can be serialized and integrated in sequence through a time series fusion strategy (such as introducing a recurrent neural network, an attention mechanism, etc. at the node or full graph level), so as to capture the changing laws of the association relationship between accounts in time evolution. For scenarios at the node level, after each snapshot graph outputs node features, a time series attention or gating mechanism can be introduced to fuse the embedding vectors of the same node in multiple time periods, thereby generating a more comprehensive node time series feature; in scenarios where a global graph representation is required, each snapshot graph can be aggregated (such as global average pooling or graph-level embedding) to obtain a single vector, and then this series of vectors can be fused in the time series dimension to generate a dynamic global representation. Regardless of whether the embedding method is adopted at the node or full graph level, the features obtained after time series fusion can be combined with the node feature representation output by the graph convolutional neural network, and finally input into the subsequent classifier or clustering model to achieve the classification (suspicious and normal) or clustering (abnormal dense subgraph) of the hype account. In this way, not only the ability of the graph neural network to depict the association structure of accounts at a single moment is retained, but also the dynamic information of different time periods is fully utilized, providing more complete and more time-related feature support for the detection of multi-account collusion hype behavior.

[0190] For further information, please refer to Figure 2 The system also includes a regulatory measures application module, which is used to simulate different regulatory strategies to adjust regulatory parameters (detection probability, penalty intensity, transaction volume upper threshold, information disclosure delay, etc.) to intervene in the market and curb speculation. By introducing the intervention mechanism of the regulatory agency, the effectiveness of various regulatory measures is evaluated. The main regulatory measures include:

[0191] 1) Improve supervision and law enforcement: Increase detection probability , increase the intensity of punishment , so that colluding speculators consider the risk of being detected when making decisions and adjust their expected returns ,Right now:

[0192]

[0193] 2) Implement transaction limits and large transaction reporting system: set a transaction volume cap , when the transaction volume exceeds the threshold, a simulated reporting mechanism is established to make it easier for regulators to detect abnormal transactions;

[0194] 3) Improve market transparency and information disclosure: Reduce information disclosure delays , public position information, and reduce the concealment of speculators and related accounts;

[0195] 4) Strengthen investor education and risk warnings: adjust retail investors’ risk appetite and the probability of following the crowd, and reduce their willingness to blindly follow the crowd;

[0196] 5) Introduce circuit breakers and price limits: set price fluctuation limits , introduce a circuit breaker mechanism to prevent abnormal price fluctuations.

[0197] Furthermore, the adjustment process of the regulatory parameters includes:

[0198] Taking the regulatory parameters as variables, several regulatory parameter combinations are randomly initialized and generated within a preset range;

[0199] Market simulation is performed based on various regulatory parameter combinations, and the fitness value is calculated based on the simulation results. Several excellent regulatory parameter combinations are selected based on the fitness value; the fitness value includes transaction anomaly, price volatility or speculator profit level;

[0200] Genetic operations are performed on excellent regulatory parameter combinations to generate new regulatory parameter combinations, and the iterations are repeated until the stopping conditions are met, and the optimal regulatory parameter combination is output.

[0201] Parameters such as detection probability, penalty intensity, transaction volume upper threshold, and information disclosure delay can be used as variables to find the optimal solution by iteratively evolving different regulatory parameter combinations in a simulation environment. For ease of explanation, the regulatory parameters to be optimized can be expressed in vector form as ; And the following fitness function F(x) is used as the optimization target:

[0202]

[0203] In the formula, It is used to measure the effectiveness of the parameter combination in curbing speculation, such as the decline in the abnormal returns of colluding speculators and the reduction in illegal trading volume; Used to measure the stability of market price and trading volume fluctuations under the current parameter combination; To assess the additional costs or complexity that regulators would incur by increasing enforcement or reducing delays in disclosing information; , and is the weight coefficient, which is used to make a trade-off between different objectives.

[0204] In the process of optimizing regulatory parameters, a certain number of regulatory parameter combinations are first randomly generated within the feasible range of each component of the regulatory parameter combination x; then, each regulatory parameter combination is substituted into the market simulation model for a certain period of operation to obtain data such as the degree of abnormal trading of collusive speculators, market volatility, and regulatory costs, and then the fitness value of each regulatory parameter combination is calculated based on the data such as the degree of abnormal trading of collusive speculators, market volatility, and regulatory costs; according to the fitness value, a number of excellent regulatory parameter combinations are selected to participate in genetic operations using strategies such as roulette or tournaments; in the crossover operation, some components of the two selected regulatory parameter combinations can be exchanged to generate new regulatory parameter combinations, for example:

[0205]

[0206] In the mutation operation, small random perturbations can be made to individual components in the selected good regulatory parameter combination:

[0207]

[0208] in, is the coefficient of variation, Indicates generating random numbers in the interval [-1,1] to ensure that the parameters traverse a larger search space.

[0209] After each generation of evolution, the system will retain the excellent parameter combinations and eliminate the poor parameter combinations, repeating the above simulation-evaluation (fitness value calculation)-selection-crossover-mutation process. As the number of evolutionary generations increases, the overall fitness of the population tends to improve, gradually converging to the optimal or near-optimal regulatory parameter configuration. Through this automatic iterative evolutionary process, it is possible to efficiently explore the balance between multi-dimensional regulatory parameters without manual repeated experiments, and adaptively adjust intervention measures in the face of complex and changing market conditions, thereby effectively improving the monitoring and prevention capabilities of collusion speculation risks. The mechanism can also dynamically correct important parameters (such as detection probability or penalty intensity) in a timely manner according to market changes, so as to achieve continuous and effective suppression of multi-account collusion speculation in different market environments.

[0210] Furthermore, the system also includes a market feedback and result analysis module, which is used to collect and analyze simulation results and evaluate the effects of speculation and regulatory measures. By counting the frequency of speculation, analyzing market price volatility, and evaluating the benefits and risks of retail investors and speculators, the impact of regulatory measures on the market can be judged.

[0211] The market feedback and results analysis module focuses on the following indicators:

[0212] 1) Frequency of speculation: Count the number of times speculators speculate under different regulatory conditions;

[0213] 2) Market price volatility: assessing price fluctuations and market stability;

[0214] 3) Returns and risks of retail investors: Analyze the average returns and risk tolerance of retail investors under different conditions;

[0215] 4) Returns and risks of speculators: assess the impact of regulatory measures on speculators’ return expectations and risks;

[0216] 5) Market efficiency and fairness: Determine whether regulatory measures have improved the market environment and increased transaction fairness.

[0217] Furthermore, the simulation system in the embodiment of the present application also includes a regulatory agency agent. The regulatory agency, as an independent agent, has monitoring, identification and punishment functions, and is used to dynamically adjust regulatory strategies based on market data and agent behavior.

[0218] Furthermore, the system also includes: a dynamic risk assessment module, which is used to use a risk model based on a Bayesian network to achieve real-time and dynamic assessment of collusion speculation behavior. Specifically, the dynamic risk assessment module defines a set of nodes reflecting speculation risks in the Bayesian network, and the nodes include market volatility, historical price anomalies, transaction correlation between collusion accounts, regulatory intensity and / or information disclosure delays, and the conditional dependencies between the nodes are represented by directed edges; for each node, a conditional probability table is constructed according to the corresponding parent node, and the conditional probability table can be initialized through statistical analysis of historical transaction data or professional knowledge a priori; when the external environment or market status changes, updated data is obtained from real-time monitoring or simulation results, and the posterior probability of each node is updated based on the Bayesian theorem, and finally the collusion speculation risk level of the current market scenario is judged based on the posterior probability of each node. Among them, the update formula of the posterior probability is:

[0219]

[0220] In the formula, Represents the parameters of each node (e.g. the distribution of transaction correlation between colluding accounts), is the newly observed market data. By synchronously updating all relevant nodes in the Bayesian network, the risk distribution of each node can be adaptively corrected under different market conditions, and the overall risk level of collusion speculation can be inferred. When the posterior probability deviates significantly from the normal range at some nodes (for example, the transaction correlation and abnormal yield increase significantly at the same time), the system can judge that the speculation risk of the current market scenario tends to increase, thereby triggering warnings or further intervention. Since the Bayesian network can flexibly characterize nonlinear dependencies between nodes and gradually correct prior assumptions when new data is obtained, the risk model based on the Bayesian network can realize continuous dynamic assessment of speculation risks in multi-market scenarios, and adapt to the detection needs of collusion speculation behavior under different investor structures and market volatility environments.

[0221] This application constructs a multi-agent market model that includes retail investors, colluding speculators and their associated accounts. It can effectively simulate the trading behaviors, interactive relationships and their impact on market dynamics of various market participants, achieve a high degree of simulation of market dynamics, make up for the deficiency of traditional models in reflecting the real market fluctuation characteristics, and provide a credible simulation basis for the analysis of market speculation behavior.

[0222] By modeling the behavior of collusive speculators and their associated accounts, this application can deeply reproduce the strategy of multi-account collaborative operation in collusive speculation, identify the speculators' hidden true intentions and evaluate their market impact; the model reveals the hiddenness and complexity of speculation behavior, and provides regulators with technical means to identify and analyze complex speculation behavior.

[0223] This application embeds multiple regulatory strategies into the risk assessment mechanism, which can simulate and analyze the effectiveness of different regulatory measures (such as increasing law enforcement, implementing trading limits, etc.) in curbing market speculation; by dynamically simulating different market regulatory scenarios, it can help regulators formulate and optimize regulatory policies and improve market fairness and transparency.

[0224] This application establishes a detailed mapping of the interactive behavior of intelligent agents and market dynamic changes, revealing the complex relationships and synergies between market participants; this mapping relationship accurately reflects key indicators such as market price fluctuations and trading activity, and can provide solid data support and theoretical basis for the detection of hype behavior.

[0225] The present application adjusts the key parameters of collusive speculators and their associated accounts to perform market simulation, and generates sample data of speculation behavior of different intensities, thereby making up for the problem of low accuracy of speculation behavior detection in the prior art due to the difficulty in obtaining speculation behavior data; and the present application generates sample data of speculation behavior of different intensities, and combines statistical distribution analysis or machine learning methods to extract the significant features of speculation behavior, thereby effectively improving the ability to identify multi-account collusion operations. Compared with the traditional method that only relies on a single abnormal threshold or rule matching, the present application can achieve more flexible and precise identification of collusion speculation strategies of different intensities and types, which helps to timely discover potential coordinated manipulation risks in real regulatory scenarios. The present application overcomes the technical bottleneck of traditional methods in identifying collusion speculation behaviors with high concealment and complexity.

[0226] This application further analyzes the behavioral correlation between multiple accounts, and more precisely locates and judges the speculation behavior of colluding speculators and their associated accounts at the account level, which improves the accuracy of identifying speculation behavior, helps to respond quickly and take intervention measures to protect market order.

[0227] Through scenario simulation, this application can dynamically simulate the evolution of collusion speculation under different market conditions and test the effectiveness of response strategies. This dynamic deduction capability allows regulators to adjust strategies in a timely manner in highly volatile or low liquidity markets, and improve the pertinence and effectiveness of regulatory measures.

[0228] Please refer to Figure 3 The embodiment of the present application also provides a market collusion speculation risk monitoring simulation method, which is applied to the market collusion speculation risk monitoring simulation system in the above embodiment, and the method includes:

[0229] Step 310: Set the basic parameters of the market environment and configure the key parameters of various intelligent entities in the market.

[0230] The basic parameters include the initial market price, price volatility, time step and transaction fees. The intelligent agents include retail investors, collusive speculators and the associated accounts of collusive speculators. The key parameters of retail investors include the maximum investment period, technical belief weight, risk aversion coefficient, fundamental volatility, historical price volatility, initial wealth and follow-up probability adjustment factor; the key parameters of the associated accounts of collusive speculators include manipulation signal adjustment factor, initial wealth, order size adjustment factor and technical belief weight; the key parameters of collusive speculators include technical belief weight, speculation strategy selection factor, risk aversion coefficient, manipulation signal adjustment factor, regulatory risk adjustment factor and initial wealth.

[0231] Step 320: Generate the initial state and agent distribution of the simulated market based on the basic parameters and key parameters.

[0232] Generate an initial price sequence based on the set initial market price and price volatility to establish the initial price environment of the market; allocate funds to various agents based on the initial wealth set in the agent parameter setting module, and determine the initial trading capabilities of various agents; adjust market parameters, the number of agents and behavioral rules, and continuously iterate the simulation until the formatting characteristics of the artificial stock market (such as sharp peaks and thick tails) match the real market data, ensuring the accuracy and reliability of subsequent simulations.

[0233] Step 330: simulate the trading behavior of various intelligent entities in the market.

[0234] Simulate the trading behaviors of various intelligent agents in the market. Different types of intelligent agents have their own trading strategies and decision-making processes, and their behaviors have different impacts on market dynamics.

[0235] The retail investor behavior simulation process includes:

[0236] Determine the investment period of retail investors based on their maximum investment period and technical belief weights;

[0237] Combine the moving average of historical prices and the recent price trend, and introduce random belief errors to determine the expected price of retail investors;

[0238] Calculate the risk level faced by retail investors by combining their fundamental volatility, historical price volatility and the adjustment factor for the probability of following the trend;

[0239] Determine the absolute risk aversion coefficient of retail investors based on their technical belief weights and risk aversion coefficients;

[0240] Calculate the price at which retail investors submit their orders based on their expected price, risk level, and absolute risk aversion coefficient;

[0241] Determine the optimal wealth investment ratio for retail investors based on their expected spot yield, risk level, and absolute risk aversion coefficient;

[0242] The order size of retail investors is calculated based on their optimal wealth investment ratio, current wealth, order submission price and current position quantity.

[0243] The simulation process of the associated account behavior of colluding speculators includes: simulating the trading behavior of retail investors, adding the speculators' adjustment items to the expected price of retail investors based on the manipulation signal adjustment factor, and obtaining the expected price of the associated account of the colluding speculators; adding the additional trading volume specified by the speculators to the order size of retail investors based on the order size adjustment factor, and obtaining the order size of the associated account of the colluding speculators.

[0244] The simulation process of the collusion speculation behavior includes:

[0245] Determine the absolute risk aversion coefficient of the collusive speculators according to their technical belief weights and risk aversion coefficients;

[0246] Calculate the order submission price of the colluding speculators based on their expected price, risk level and absolute risk aversion coefficient;

[0247] Determine the optimal wealth investment ratio of the collusive speculators based on their expected spot yield, risk level and absolute risk aversion coefficient;

[0248] The order size of the colluding speculators is calculated based on their optimal wealth investment ratio, current wealth, order submission price and current position quantity.

[0249] Step 340: simulate the hype behavior of the speculator.

[0250] There are two types of speculation: single-point speculation and multi-point speculation. In single-point speculation, colluding speculators speculate independently through their own accounts without using associated accounts. The speculation process includes:

[0251] S1. Accumulation stage: speculators gradually buy at low levels to build positions and avoid attracting market attention;

[0252] S2. Pull-up stage: speculators buy with larger orders and higher prices, pushing up market prices and creating an upward trend;

[0253] S3. Selling stage: speculators gradually sell their positions at high levels to gain profits;

[0254] In multi-point speculation, speculators use their own accounts and multiple associated accounts to coordinate market speculation, increasing the concealment of speculation. The speculation process includes:

[0255] S1. Fund-accumulation stage: speculators and related accounts jointly buy at low levels to diversify their positions and avoid attracting market attention;

[0256] S2. Pull-up phase: Speculators and related accounts buy at different times, prices, and quantities in a coordinated manner to push up prices and create the illusion of an active market;

[0257] S3. Shipping stage: Speculators and related accounts sell their positions in batches to maintain price stability and avoid sudden price drops.

[0258] Step 350: Adjust key parameters of the colluding speculator and the associated account of the colluding speculator to obtain sample data of speculation behavior with different speculation intensities, and identify speculation scenarios with different speculation intensities through statistical distribution analysis methods or machine learning methods.

[0259] Adjust the key parameters of the collusive speculators and their associated accounts (such as the speculation strategy selection factor, order size adjustment factor, risk aversion coefficient, etc.), and gradually generate multiple sets of artificial market simulation data from low to high levels of speculation. These speculation scenarios with different speculation intensities can reflect the market-level changes (such as price distribution characteristics, trading volume distribution characteristics, position distribution characteristics, etc.) brought about by the collusive speculators' implementation of different levels of speculation in the market. Specifically, the speculation intensity can be controlled by parameter adjustment to obtain speculation sample data with different speculation intensities; in each simulation, for the collusive speculators and their associated accounts, the key parameters such as the manipulation signal adjustment factor, order size adjustment factor, and speculation strategy selection factor are successively increased or decreased, so that they can change from "almost no intervention in the market" to "frequent and large-scale intervention in the market". Subsequently, based on these multi-level and multi-dimensional market indicators, hype behaviors of different intensities can be identified in the distribution or distance space. For example, the distribution differences in market indicators of various hype scenarios, such as non-hype / weak hype scenarios, medium hype scenarios, and strong hype scenarios, can be compared through distribution difference measurement, and statistical distance measurement (such as KL divergence, JS divergence, Wasserstein distance, etc.) can be used to judge the degree of deviation caused by hype to the overall market distribution; or after mapping the market indicators of each scenario to a feature vector, clustering methods (such as K-means, DBSCAN, etc.) or classification methods (such as SVM, random forest, etc.) can be used to distinguish non-hype / weak hype scenarios, medium hype scenarios, and strong hype scenarios. Through the above-mentioned method, on the one hand, artificial market simulation is used to generate multi-level speculation behavior data, and on the other hand, statistical distribution analysis or machine learning methods are combined to extract the significant features of speculation behavior, thereby effectively improving the ability to identify multi-account collusion operations; compared with the traditional method that only relies on a single abnormal threshold or rule matching, this embodiment can achieve more flexible and precise identification of collusion speculation strategies of different intensities and types, which is helpful to timely discover potential collaborative manipulation risks in real regulatory scenarios.

[0260] Furthermore, by adjusting the order size, order time distribution and position change rules of each account in the simulated market, account data samples with different speculation characteristics can be obtained;

[0261] The feature representation of each account is extracted from the account data sample through a graph neural network, and the feature representation of each account is classified or clustered through a classifier to obtain the collusion speculation account;

[0262] Or, quantitatively describe account data samples in different feature dimensions to construct account portraits and extract feature indicators for distinguishing normal transactions from suspicious transactions. Feature dimensions include transaction frequency, buying and selling sequence, and correlation with market price fluctuations;

[0263] Conduct graph association analysis, correlation measurement statistics, or machine learning model classification on feature indicators to identify abnormal linkage relationships between accounts and obtain accounts that collude to speculate.

[0264] The market collusion speculation risk monitoring simulation method provided by the present application, by constructing a multi-agent model including retail investors, collusion speculators and their associated accounts, deeply simulates the decentralized trading and synchronous operation strategies of collusion speculators, accurately captures their hidden true intentions, and helps to improve the detection accuracy of collusion speculation behavior; by adjusting the key parameters of collusion speculators and their associated accounts for market simulation, speculative behavior sample data of different speculation intensities are generated, which makes up for the problem of low accuracy of speculative behavior detection caused by the difficulty in obtaining speculative behavior data in the prior art. In addition, the present application generates speculative behavior sample data of different intensities, and combines statistical distribution analysis or machine learning methods to extract the significant features of speculative behavior, thereby effectively improving the ability to identify multi-account collusion operations. Compared with the traditional method that only relies on a single abnormal threshold or rule matching, the present application can achieve more flexible and precise identification of collusion speculation strategies of different intensities and types, which helps to timely discover potential coordinated manipulation risks in real regulatory scenarios.

[0265] Furthermore, step 350 also includes step 360, simulating different regulatory strategies to adjust regulatory parameters to intervene in the market and curb speculation.

[0266] By introducing the intervention mechanism of the regulatory agency, the effectiveness of various regulatory measures is evaluated. The main regulatory measures include:

[0267] 1) Improve supervision and law enforcement: Increase detection probability , increase the intensity of punishment , so that colluding speculators consider the risk of being detected when making decisions and adjust their expected returns ,Right now:

[0268]

[0269] 2) Implement transaction limits and large transaction reporting system: set a transaction volume cap , when the transaction volume exceeds the threshold, a simulated reporting mechanism is established to make it easier for regulators to detect abnormal transactions;

[0270] 3) Improve market transparency and information disclosure: Reduce information disclosure delays , public position information, and reduce the concealment of speculators and related accounts;

[0271] 4) Strengthen investor education and risk warnings: adjust retail investors’ risk appetite and the probability of following the crowd, and reduce their willingness to blindly follow the crowd;

[0272] 5) Introduce circuit breakers and price limits: set price fluctuation limits , introduce a circuit breaker mechanism to prevent abnormal price fluctuations.

[0273] Furthermore, the adjustment process of the regulatory parameters includes:

[0274] Taking the regulatory parameters as variables, several regulatory parameter combinations are randomly initialized and generated within a preset range;

[0275] Market simulation is performed based on various regulatory parameter combinations, and the fitness value is calculated based on the simulation results. Several excellent regulatory parameter combinations are selected based on the fitness value; the fitness value includes transaction anomaly, price volatility or speculator profit level;

[0276] Genetic operations are performed on excellent regulatory parameter combinations to generate new regulatory parameter combinations, and the iterations are repeated until the stopping conditions are met, and the optimal regulatory parameter combination is output.

[0277] Furthermore, the method also includes step 370, collecting and analyzing simulation results to evaluate the effects of hype and regulatory measures.

[0278] By counting the frequency of speculative behavior, analyzing market price volatility, and evaluating the returns and risks of retail investors and speculators, we can judge the impact of regulatory measures on the market.

[0279] The market feedback and results analysis module focuses on the following indicators:

[0280] 1) Frequency of speculation: Count the number of times speculators speculate under different regulatory conditions;

[0281] 2) Market price volatility: assessing price fluctuations and market stability;

[0282] 3) Returns and risks of retail investors: Analyze the average returns and risk tolerance of retail investors under different conditions;

[0283] 4) Returns and risks of speculators: assess the impact of regulatory measures on speculators’ return expectations and risks;

[0284] 5) Market efficiency and fairness: Determine whether regulatory measures have improved the market environment and increased transaction fairness.

[0285] Furthermore, the method also includes step 380, which uses a risk model based on a Bayesian network to conduct a real-time and dynamic evaluation of the collusion speculation behavior.

[0286] Specifically, the dynamic risk assessment module defines a set of nodes reflecting speculation risk in the Bayesian network, including market volatility, historical price anomaly, transaction correlation between collusion accounts, regulatory intensity and / or information disclosure delay, and the conditional dependency between nodes is represented by directed edges; for each node, a conditional probability table is constructed according to the corresponding parent node, and the conditional probability table can be initialized by statistical analysis of historical transaction data or professional knowledge prior; when the external environment or market status changes, updated data is obtained from real-time monitoring or simulation results, and the posterior probability of each node is updated based on the Bayesian theorem, and finally the collusion speculation risk level of the current market scenario is judged according to the posterior probability of each node. By synchronously updating all relevant nodes in the Bayesian network, the risk distribution of each node can be adaptively corrected under different market conditions, and the overall collusion speculation risk level can be inferred. When the posterior probability deviates significantly from the normal range at some nodes (for example, the transaction correlation and abnormal yield rate increase significantly at the same time), the system can judge that the speculation risk of the current market scenario tends to increase, thereby triggering warnings or further intervention. Since Bayesian networks can flexibly characterize nonlinear dependencies between nodes and gradually correct prior assumptions when new data is obtained, in multi-market scenarios, risk models based on Bayesian networks can achieve continuous dynamic assessment of speculation risks and adapt to the needs of detecting collusion speculation behavior under different investor structures and market volatility environments.

[0287] The embodiment of the present application also provides an electronic device, the device comprising a processor and a memory;

[0288] The memory is used to store the program code and transmit the program code to the processor;

[0289] The processor is used to execute the market collusion speculation risk monitoring simulation method in the aforementioned method embodiment according to the instructions in the program code.

[0290] The embodiment of the present application also provides a computer-readable storage medium, which is used to store program code. When the program code is executed by a processor, the market collusion and speculation risk monitoring simulation method in the aforementioned method embodiment is implemented.

[0291] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific process of the method described above can refer to the corresponding process in the aforementioned system embodiment, and will not be repeated here.

[0292] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0293] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0294] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0295] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0296] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0297] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions for executing all or part of the steps of the method described in each embodiment of the present application through a computer device (which can be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (full name in English: Read-Only Memory, English abbreviation: ROM), random access memory (full name in English: Random Access Memory, English abbreviation: RAM), disk or optical disk and other media that can store program codes.

[0298] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A market collusion speculation risk monitoring simulation system, characterized in that: include: A system parameter setting module, used to set basic parameters of the market environment, including initial market price, price volatility, time step and transaction fee; An agent parameter setting module, used to configure key parameters of various agents in the market, including retail investors, colluding speculators, and associated accounts of colluding speculators; A market initialization and calibration module, used to generate the initial state and agent distribution of the simulated market based on the basic parameters and the key parameters; Agent behavior simulation module, used to simulate the trading behavior of various agents in the market; The module for implementing the collusion hype strategy is used to simulate the hype behavior of the hype man; The hype identification module is used to adjust the key parameters of the colluding hype person and the associated account of the colluding hype person to obtain hype behavior sample data of different hype intensities, and identify hype scenarios of different hype intensities through statistical distribution analysis methods or machine learning methods.

2. The market collusion speculation risk monitoring simulation system according to claim 1 is characterized in that: The key parameters of retail investors include maximum investment period, technical belief weight, risk aversion coefficient, fundamental volatility, historical price volatility, initial wealth, and follow-up probability adjustment factor; The key parameters of the associated accounts of the collusive speculators include the manipulation signal adjustment factor, initial wealth, order size adjustment factor, and technical belief weight; The key parameters of the collusive speculators include technical belief weight, speculation strategy selection factor, risk aversion coefficient, manipulation signal adjustment factor, regulatory risk adjustment factor and initial wealth.

3. The market collusion speculation risk monitoring simulation system according to claim 2 is characterized in that: The market initialization and calibration module is specifically used for: Generate an initial price sequence according to the initial market price and the price volatility to establish an initial price environment of the market; Allocate funds to various intelligent entities according to the initial wealth, and determine the initial trading capacity of each intelligent entity; Adjust market parameters, number of agents, and behavioral rules through iterative simulation until the formatting characteristics of the artificial stock market match real market data.

4. The market collusion speculation risk monitoring simulation system according to claim 2 is characterized in that: The agent behavior simulation module includes a retail investor behavior simulation submodule, and the retail investor behavior simulation submodule is used to: Determining the investment period of the retail investor based on the maximum investment period of the retail investor and the technical belief weight; Combining the moving average of historical prices and the recent price trend, and introducing random belief errors, determine the expected price of the retail investors; Calculate the risk level faced by the retail investors by combining the fundamental volatility, historical price volatility and the trend-following probability adjustment factor of the retail investors; Determining the absolute risk aversion coefficient of the retail investor based on the technical belief weight and risk aversion coefficient of the retail investor; Calculating the order submission price of the retail investor based on the retail investor's expected price, risk level and absolute risk aversion coefficient; Determining the optimal wealth investment ratio of the retail investors based on the expected spot yield, risk level faced and absolute risk aversion coefficient of the retail investors; The order size of the retail investor is calculated based on the retail investor's optimal wealth investment ratio, current wealth, order submission price and current position quantity.

5. The market collusion speculation risk monitoring simulation system according to claim 2 is characterized in that: The agent behavior simulation module includes a submodule for simulating the behavior of the associated account of the colluding speculator, and the submodule for simulating the behavior of the associated account of the colluding speculator is used to: Simulating the trading behavior of the retail investors, adding the adjustment item of the speculators to the expected price of the retail investors based on the manipulation signal adjustment factor, and obtaining the expected price of the associated account of the colluding speculators; Based on the order size adjustment factor, the additional transaction volume specified by the speculator is increased on the basis of the order size of the retail investor to obtain the order size of the associated account of the colluding speculator.

6. The market collusion speculation risk monitoring simulation system according to claim 2 is characterized in that: The intelligent agent behavior simulation module includes a collusion speculator behavior simulation submodule, and the collusion speculator behavior simulation submodule is used to: Determine the absolute risk aversion coefficient of the colluding speculators according to the technical belief weight and risk aversion coefficient of the colluding speculators; Calculate the order submission price of the colluding speculators according to the expected price, risk level faced and absolute risk aversion coefficient of the colluding speculators; Determine the optimal wealth investment ratio of the colluding speculators according to the expected spot rate of return, the risk level faced and the absolute risk aversion coefficient of the colluding speculators; The order size of the colluding speculator is calculated based on the colluding speculator's optimal wealth investment ratio, current wealth, order submission price and current position quantity.

7. The market collusion speculation risk monitoring simulation system according to claim 2 is characterized in that: The system also includes: a regulatory measures application module, which is used to simulate different regulatory strategies to adjust regulatory parameters in order to intervene in the market and curb speculation; Specifically, it is used to increase regulatory risk adjustment factors and penalties, so that speculators consider the detected risks when making decisions and adjust the corresponding expected returns; Set a transaction volume cap. When the transaction volume exceeds the threshold, the simulation reporting mechanism generates a report; Reduce information disclosure delays and make position information public; Adjust the risk appetite and follow-up probability of the retail investors; Set price fluctuation limits and introduce circuit breaker mechanisms.

8. The market collusion speculation risk monitoring simulation system according to claim 7 is characterized in that: The adjustment process of the regulatory parameters includes: Taking the regulatory parameters as variables, randomly initializing and generating a plurality of regulatory parameter combinations within a preset range; Conduct market simulation based on each of the regulatory parameter combinations, calculate fitness values ​​based on the simulation results, and select several excellent regulatory parameter combinations based on the fitness values; the fitness values ​​include transaction anomaly, price volatility, or speculator profit level; Genetic operations are performed on the excellent regulatory parameter combinations to generate new regulatory parameter combinations, and the iterations are repeated until a stop condition is met, and the optimal regulatory parameter combination is output.

9. The market collusion speculation risk monitoring simulation system according to claim 2 is characterized in that: The system also includes a market feedback and results analysis module, which is used to collect and analyze simulation results and evaluate the effects of speculation and regulatory measures; It is specifically used to count the number of times speculators speculate under different regulatory conditions; Assess price volatility and market stability; Analyze the average returns and risk tolerance of retail investors under different conditions; Assess the impact of regulatory measures on speculators’ profit expectations and risks; Determine whether regulatory measures have improved the market environment.

10. The market collusion speculation risk monitoring simulation system according to claim 1 is characterized in that: The system also includes a regulatory agency agent, which is used to dynamically adjust regulatory strategies based on market data and the behavior of each agent.

11. The market collusion speculation risk monitoring simulation system according to claim 1 is characterized in that: The hype identification module is specifically used to: Adjust the key parameters of the colluding speculators and the associated accounts of the colluding speculators to obtain sample data of speculation behaviors with different speculation intensities, compare the distribution differences of speculation scenarios with different speculation intensities on market indicators through distribution difference measurement, and use statistical distance measurement to determine the degree of deviation caused by the speculation behaviors to the overall market distribution, so as to identify speculation scenarios with different speculation intensities; Alternatively, the market indicators of the sample data of the hype behavior at each hype intensity are mapped to feature vectors, and then the hype scenarios of different hype intensities are identified through a clustering model or a classification model.

12. The market collusion speculation risk monitoring simulation system according to claim 1 is characterized in that: The speculation identification module is also used to obtain account data samples with different speculation characteristics by adjusting the order size, order time distribution and position change rules of each account in the simulated market; Extracting feature representations of each account from the account data sample through a graph neural network, classifying the feature representations of each account through a classifier or clustering the feature representations of each account through a clustering model, and obtaining collusion speculation accounts; Or, quantitatively describe the account data samples in different feature dimensions to construct an account profile and extract feature indicators for distinguishing normal transactions from suspicious transactions. The feature dimensions include transaction frequency, buying and selling sequence, and correlation with market price fluctuations; Perform graph association analysis, correlation measurement statistics or machine learning model classification on the characteristic indicators to identify abnormal linkage relationships between accounts and obtain collusion speculation accounts.

13. The market collusion speculation risk monitoring simulation system according to claim 12 is characterized in that: The training process of the graph neural network includes: Take the accounts as nodes, construct edges based on the back-to-back transactions and fund flows between accounts, and attach the position, order frequency, transaction amount and / or correlation to each node and its edge to obtain a graph structure; The association representation between accounts is learned based on node and edge features through graph convolutional networks.

14. The market collusion speculation risk monitoring simulation system according to claim 1 is characterized in that: The system also includes: a dynamic risk assessment module, which is used to: A set of nodes reflecting speculation risk is defined in the Bayesian network. The nodes include market volatility, historical price anomaly, transaction correlation between collusion accounts, regulatory intensity and / or information disclosure delay. The conditional dependency relationship between the nodes is represented by directed edges. For each node, a conditional probability table is constructed based on the corresponding parent node; When the external environment or market status changes, updated data is obtained from real-time monitoring or simulation results, and the posterior probability of each node is updated based on Bayesian theorem; The risk level of collusion speculation in the current market scenario is determined based on the posterior probability of each node.

15. A market collusion speculation risk monitoring simulation method, characterized in that: include: Setting the basic parameters of the market environment and configuring the key parameters of various intelligent agents in the market, the basic parameters include the initial market price, price volatility, time step and transaction fees, and the intelligent agents include retail investors, colluding speculators and the associated accounts of colluding speculators; Generate the initial state and agent distribution of the simulated market based on the basic parameters and the key parameters; Simulate the trading behaviors of various intelligent agents in the market; Simulate the hype behavior of speculators; The key parameters of the colluding speculator and the associated account of the colluding speculator are adjusted to obtain sample data of speculation behavior with different speculation intensities, and the speculation scenarios with different speculation intensities are identified through statistical distribution analysis methods or machine learning methods.

16. An electronic device, characterized in that: The device comprises a processor and a memory; The memory is used to store program codes and transmit the program codes to the processor; The processor is used to execute the market collusion speculation risk monitoring simulation method described in claim 15 according to the instructions in the program code.

17. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store program code, and when the program code is executed by the processor, it implements the market collusion speculation risk monitoring simulation method described in claim 15.

Images