Artificial intelligence-based data generation method and apparatus, computer device, and medium

Abstract

The application belongs to the technical field of artificial intelligence, and relates to a data generation method and device based on artificial intelligence, computer equipment and a medium, which comprises the following steps: collecting historical market data corresponding to a target product, and acquiring definition data corresponding to the historical market data; performing strategy construction based on the historical market data and the definition data to generate an initial strategy pool; performing optimization processing on the initial strategy pool based on an evolution algorithm to obtain a first strategy; performing strategy updating processing on the first strategy based on real-time market data corresponding to the target product by using a reinforcement learning algorithm to obtain a second strategy; acquiring output action results corresponding to the real-time market data generated based on each second strategy; performing fusion processing on the output action results based on a fusion mode to generate a target processing strategy; and performing output processing on the target processing strategy. The application can be applied to a strategy generation scene in the field of financial technology, and can effectively improve the processing efficiency and flexibility of strategy generation.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology and can be applied to the financial technology field, particularly to data generation methods, devices, computer equipment, and storage media based on artificial intelligence. Background Technology

[0002] Traditional investment trading strategy generation methods commonly employ fixed-rule model architectures, such as constructing trading trigger conditions through preset technical indicators (e.g., moving average crossovers). These strategies heavily rely on human experience for parameter tuning during the design phase and require backtesting with historical data to verify their effectiveness. However, the efficiency of strategy generation is often low due to the trial-and-error costs of manual parameter tuning and the computational resources consumed by backtesting. Furthermore, traditional strategies typically maintain static, fixed rule logic and parameter configurations after deployment, lacking dynamic adjustment capabilities, thus resulting in low flexibility in trading strategy generation.

[0003] Therefore, there is an urgent need to develop an automated implementation method that can effectively improve the efficiency and flexibility of transaction processing strategy generation, thereby providing financial institutions with more reliable transaction support in complex market environments. Summary of the Invention

[0004] The purpose of this application is to propose a data generation method, apparatus, computer device, and storage medium based on artificial intelligence, so as to solve the technical problems of low efficiency and low flexibility in existing transaction processing strategy generation methods.

[0005] Firstly, an artificial intelligence-based data generation method is provided, including: Collect historical market data corresponding to the target product, and obtain definition data corresponding to the historical market data; Based on the historical market data and the defined data, strategy construction processing is performed to generate a corresponding initial strategy pool; The initial policy pool is optimized based on a preset evolutionary algorithm to select the corresponding first policy; wherein, the number of the first policies includes multiple policies. Collect real-time market data corresponding to the target product; Based on the real-time market data, a preset reinforcement learning algorithm is used to update the first strategy to obtain the corresponding second strategy. Obtain the output action results corresponding to the real-time market data generated based on each of the second strategies; The output action results are fused based on a preset fusion method to generate a corresponding target processing strategy; The target processing strategy is then processed for output.

[0006] Secondly, an artificial intelligence-based data generation device is provided, comprising: The first processing module is used to collect historical market data corresponding to the target product and obtain definition data corresponding to the historical market data; The generation module is used to perform strategy construction processing based on the historical market data and the defined data to generate a corresponding initial strategy pool. An optimization module is used to optimize the initial policy pool based on a preset evolutionary algorithm to select a corresponding first policy; wherein the number of the first policies includes multiple policies. The collection module is used to collect real-time market data corresponding to the target product; The update module is used to update the first strategy based on the real-time market data using a preset reinforcement learning algorithm to obtain the corresponding second strategy. The acquisition module is used to acquire the output action results corresponding to the real-time market data generated based on each of the second strategies; The fusion module is used to perform fusion processing on the output action results based on a preset fusion method to generate a corresponding target processing strategy; The output module is used to perform output processing on the target processing strategy.

[0007] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described artificial intelligence-based data generation method.

[0008] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the aforementioned artificial intelligence-based data generation method.

[0009] In the above-mentioned solution implemented by the data generation method, apparatus, computer equipment, and storage medium based on artificial intelligence, historical market data corresponding to the target product is first collected, and definition data corresponding to the historical market data is obtained; then, strategy construction processing is performed based on the historical market data and the definition data to generate a corresponding initial strategy pool; subsequently, the initial strategy pool is optimized based on a preset evolutionary algorithm to filter and obtain a corresponding first strategy; wherein, the number of the first strategies includes multiple; subsequently, real-time market data corresponding to the target product is collected; and based on the real-time market data, a preset reinforcement learning algorithm is used to perform strategy update processing on the first strategies to obtain a corresponding second strategy; further, output action results corresponding to the real-time market data generated based on each second strategy are obtained; in another step, the output action results are fused based on a preset fusion method to generate a corresponding target processing strategy; finally, the target processing strategy is output processed. Based on the above automated processing flow, this application generates an initial strategy pool by constructing strategies using collected historical market data and acquired definition data. Then, it optimizes the initial strategy pool using an evolutionary algorithm to select a first strategy. Next, based on collected real-time market data, it updates the first strategy using a reinforcement learning algorithm to obtain a second strategy. It then obtains the output action results corresponding to the real-time market data generated by each second strategy. Finally, it fuses the output action results using a fusion method to generate and output the final target processing strategy. Thus, this application first generates an initial strategy pool based on historical market data, optimizes it using an evolutionary algorithm, and then performs real-time decision-making and strategy updates based on actual market data, using a combination of reinforcement learning algorithms and fusion methods. This achieves intelligent and accurate automatic generation of strategies related to the target product, effectively improving the processing efficiency and flexibility of strategy generation. Attached Figure Description

[0010] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is an exemplary system architecture diagram to which this application can be applied; Figure 2 This is a flowchart of an embodiment of the artificial intelligence-based data generation method according to this application; Figure 3 This is a schematic diagram of a structure of an embodiment of the artificial intelligence-based data generation apparatus according to this application; Figure 4 This is a schematic diagram of the structure of one embodiment of the computer device according to this application. Detailed Implementation

[0012] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0013] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0014] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0015] like Figure 1 As shown, system architecture 100 may include terminal device 101, network 102, and server 103. Terminal device 101 may be a laptop 1011, tablet 1012, or mobile phone 1013. Network 102 is used as a medium to provide a communication link between terminal device 101 and server 103. Network 102 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0016] Users can use terminal device 101 to interact with server 103 via network 102 to receive or send messages, etc. Various communication client applications can be installed on terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0017] Terminal device 101 can be various electronic devices with a display screen and support web browsing. In addition to laptops 1011, tablets 1012, or mobile phones 1013, terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer, and a desktop computer, etc.

[0018] Server 103 can be a server that provides various services, such as a backend server that provides support for the pages displayed on terminal device 101.

[0019] It should be noted that the AI-based data generation method provided in this application is generally executed by a server / terminal device, and correspondingly, the AI-based data generation device is generally located in the server / terminal device.

[0020] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0021] Continue to refer to Figure 2 This document illustrates a flowchart of an embodiment of the AI-based data generation method according to this application. The order of steps in the flowchart can be changed, and some steps can be omitted, depending on different needs. The AI-based data generation method provided in this application can be applied to any scenario requiring strategy generation, and therefore can be applied to products in these scenarios, such as strategy generation products in the fintech field. The AI-based data generation method includes the following steps: Step S201: Collect historical market data corresponding to the target product and obtain definition data corresponding to the historical market data.

[0022] In this embodiment, the artificial intelligence-based data generation method operates on an electronic device (e.g., Figure 1The server / terminal device shown can acquire historical market data corresponding to the target product via wired or wireless connection. It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra-wideband) connections, and other currently known or future-developed wireless connection methods. The executing entity of this application is specifically a data generation system, also known as a trading system, which can be simply referred to as the system. This application can be applied to scenarios in the fintech field where processing strategies corresponding to investment products are generated. The aforementioned target product is an investment product with market-based trading data, such as stocks, funds, bonds, futures, cryptocurrencies, etc.

[0023] This can be achieved by obtaining historical market data (including price, trading volume, technical indicators, etc.) related to the target products from multiple reliable market data sources (such as exchange APIs, financial data platforms, etc.), and by comprehensively examining the collected historical market data to handle missing and outlier values, thereby improving data quality. Regarding price... It represents the market price at time t and is a key indicator reflecting asset value; trading volume The number of transactions at time t reflects the market's activity level. Various technical indicators... The calculation of these metrics is based on common financial analysis methods. For example, the moving average (MA) is obtained by averaging prices over a certain time window. The Relative Strength Index (RSI), on the other hand, measures the overbought or oversold condition of the market by comparing the magnitude of price increases and decreases over a certain period.

[0024] In addition, the aforementioned data definitions include state definitions, action definitions, and reward function definitions. State definitions involve organizing the collected historical market data by time t to construct the market state. This vector comprehensively describes the overall market situation at time t, providing a foundation for subsequent trading decisions. The action definition clearly defines the set of trading actions. Here, -1 represents selling, meaning the investor believes the market price will fall and sells assets to reduce losses or gain profits; 0 represents holding, meaning the investor believes the current market situation is uncertain and chooses to maintain the current asset status; +1 represents buying, indicating the investor expects the market price to rise and buys assets in hopes of future price increases. Reward function definition. middle, It is the price change at time t, that is = It reflects the price fluctuation of an asset at time t; These are transaction costs, including fees such as commissions. Used to measure the magnitude of changes in trading actions, when trading actions change frequently, transaction costs have a significant impact on returns. The reward function comprehensively considers the returns brought by price changes and the impact of transaction costs on returns, in order to guide trading strategies to make more rational decisions.

[0025] Furthermore, the goal of the plan is to maximize the cumulative expected return, where The discount factor is calculated using the following formula: By collecting and organizing market data, constructing market states, defining trading actions and reward functions, a clear environment and objectives are provided for subsequent strategy generation and optimization, enabling subsequent algorithms to operate and be evaluated within a unified framework.

[0026] Step S202: Based on the historical market data and the defined data, perform strategy construction processing to generate the corresponding initial strategy pool.

[0027] In this embodiment, the system trains multiple basic strategies (which can be composed of neural networks or rule models) using historical market data and the aforementioned defined data to form a corresponding initial strategy pool. Specifically, during model training, taking a neural network as an example, the market state St is used as input, and the probability distribution of trading actions is used as output. The neural network continuously adjusts its parameters using a backpropagation algorithm. The backpropagation algorithm calculates the gradient layer by layer from the output layer to the input layer based on the error between the output result and the true optimal action, and updates the network parameters accordingly. This allows the action probability distribution output by the model to better fit the optimal action in historical transaction data. After training, several basic strategies are obtained. Each strategy Output in state Action probability distribution ,in, For transaction actions, It is determined by parameters A defined neural network function.

[0028] Using multiple models to train the basic policy increases the diversity of policies, enabling the subsequent optimization process to explore the optimal policy space from multiple different perspectives.

[0029] Step S203: Optimize the initial policy pool based on a preset evolutionary algorithm to select the corresponding first policy; wherein, the number of the first policies includes multiple policies.

[0030] In this embodiment, the specific implementation process of optimizing the initial policy pool based on the preset evolutionary algorithm to select the corresponding first policy will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0031] Step S204: Collect real-time market data corresponding to the target product.

[0032] In this embodiment, new market data (i.e., real-time market data) is continuously collected during real-time trading. Specifically, this can be achieved by establishing efficient data acquisition tools to obtain data from various data sources in real time and transmit it to the system promptly. For example, market data for a target product, such as a stock, can be obtained in real time by connecting to a data interface of a stock exchange.

[0033] Step S205: Based on the real-time market data, use a preset reinforcement learning algorithm to update the first strategy to obtain the corresponding second strategy.

[0034] In this embodiment, the specific implementation process of updating the first strategy using a preset reinforcement learning algorithm based on the real-time market data to obtain the corresponding second strategy will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0035] Step S206: Obtain the output action results corresponding to the real-time market data generated based on each of the second strategies.

[0036] In this embodiment, each of the above-mentioned second strategies will output a corresponding action probability distribution (including selling probability, holding probability, and buying probability) based on the current real-time market data, that is, the above-mentioned output action result.

[0037] Step S207: The output action result is fused based on a preset fusion method to generate a corresponding target processing strategy.

[0038] In this embodiment, the specific implementation process of fusing the output action results based on the preset fusion method to generate the corresponding target processing strategy will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0039] Step S208: Output processing is performed on the target processing strategy.

[0040] In this embodiment, the generated target processing strategy can be sent to the corresponding user via email, message, or interface display, thereby completing the output processing of the generated target processing strategy. Additionally, actual trading decisions can be made on the target product based on the obtained target processing strategy. For example, the generated target processing strategy can be sent to the trading system to execute buy, sell, or hold operations. Simultaneously, relevant information regarding the trading decision processing, such as trading time, trading instrument, trading quantity, and trading price, is recorded for subsequent strategy evaluation and auditing.

[0041] This application first collects historical market data corresponding to the target product and obtains definition data corresponding to the historical market data; then, it performs strategy construction processing based on the historical market data and the definition data to generate a corresponding initial strategy pool; subsequently, it optimizes the initial strategy pool based on a preset evolutionary algorithm to select a corresponding first strategy; wherein, the number of first strategies includes multiple strategies; next, it collects real-time market data corresponding to the target product; and based on the real-time market data, it uses a preset reinforcement learning algorithm to perform strategy update processing on the first strategies to obtain a corresponding second strategy; further, it obtains the output action results corresponding to the real-time market data generated based on each second strategy; then, it performs fusion processing on the output action results based on a preset fusion method to generate a corresponding target processing strategy; finally, it performs output processing on the target processing strategy. Based on the above automated processing flow, this application generates an initial strategy pool by constructing strategies using collected historical market data and acquired definition data. Then, it optimizes the initial strategy pool using an evolutionary algorithm to select a first strategy. Next, based on collected real-time market data, it updates the first strategy using a reinforcement learning algorithm to obtain a second strategy. It then obtains the output action results corresponding to the real-time market data generated by each second strategy. Finally, it fuses the output action results using a fusion method to generate and output the final target processing strategy. Thus, this application first generates an initial strategy pool based on historical market data, optimizes it using an evolutionary algorithm, and then performs real-time decision-making and strategy updates based on actual market data, using a combination of reinforcement learning algorithms and fusion methods. This achieves intelligent and accurate automatic generation of strategies related to the target product, effectively improving the processing efficiency and flexibility of strategy generation.

[0042] In some alternative implementations, step S203 includes the following steps: For each strategy in the initial strategy pool, the fitness corresponding to each strategy is calculated using the historical market data.

[0043] In this embodiment, the fitness calculation process is performed for each policy in the initial policy pool. (No. One strategy), using historical market data to calculate its fitness. The fitness function comprehensively considers the risk-reward ratio and risk control, and may specifically include: By using the Sharpe ratio The risk-reward ratio measures the excess return earned per unit of risk. Maximum drawdown. Constraining and controlling risk, maximum drawdown This indicates the percentage decrease in assets from their highest to their lowest point within a certain period. It is the weighting coefficient for the maximum drawdown, used to balance the relationship between returns and risks.

[0044] Based on the fitness, the policy is subjected to selection, crossover, and mutation operations to obtain a newly generated third policy.

[0045] In this embodiment, the selection operation sorts the strategies according to their fitness values ​​and selects the strategy with the higher fitness as the parent strategy. The selection method can be roulette wheel selection, where the probability of each strategy being selected is proportional to its fitness value; or tournament selection, where a certain number of strategies are randomly selected for comparison, and the strategy with the highest fitness is selected as the parent strategy. Then, the crossover operation randomly selects two strategies from the selected parent strategies. and New strategy parameters are generated according to certain crossover probabilities and crossover patterns. For example, using a single-point crossover, a random crossover point is selected, and the parameters of the two parent policies after that point are swapped to obtain the new policy parameters. .in, It's the cross weight, which controls the degree of fusion between the parameters of the two parent policies. This is random noise, used to increase policy diversity and avoid getting trapped in local optima. Subsequently, the mutation operation affects the newly generated policy parameters. Mutation operations are performed according to a certain mutation probability, such as randomly changing the values ​​of certain parameters, like adding a small random perturbation to a parameter, to further explore the policy space and generate a completely new set of policies, namely the third policy (which may have the same number as or more than the original policy pool, depending on the specific implementation).

[0046] Calculate the first fitness of the third strategy.

[0047] In this embodiment, the process of calculating the first fitness of the third strategy can be referred to the aforementioned process of calculating the fitness corresponding to each strategy using the historical market data, and will not be elaborated further here.

[0048] The strategy and the third strategy are integrated to obtain a corresponding strategy set.

[0049] In this embodiment, a temporary strategy set can be formed by integrating the above strategy with the generated third strategy.

[0050] Based on the fitness and the first fitness, the strategy set is sorted in descending order of value to obtain the corresponding sorting result.

[0051] In this embodiment, the strategy set is sorted in descending order of fitness value to obtain the corresponding ranking result. A higher fitness value indicates a better strategy (because the goal is to minimize risk-adjusted drawdown while maximizing the risk-reward ratio).

[0052] A specified number of fourth strategies are selected from the sorting results, and the fourth strategies are used as the first strategy.

[0053] In this embodiment, the selection of the specified number K is not specifically limited and can be determined according to actual business needs. The top K optimal strategies in the strategy set, i.e., the fourth strategy, are retained and used as the corresponding first strategy.

[0054] This application calculates the fitness of each strategy in the initial strategy pool using historical market data; then, based on the fitness, performs selection, crossover, and mutation operations on the strategies to obtain a newly generated third strategy; and calculates the first fitness of the third strategy; subsequently, it integrates the strategies with the third strategy to obtain a corresponding strategy set; then, based on the fitness and the first fitness, it sorts the strategy set in descending order of value to obtain a corresponding sorting result; finally, it selects a specified number of fourth strategies from the sorting result and uses these fourth strategies as the first strategy. Based on the above processing flow, this application uses an evolutionary algorithm to optimize the strategy pool by simulating the process of biological evolution, performing selection, crossover, and mutation operations on the strategies in the initial strategy pool, continuously selecting strategies with higher fitness, and gradually optimizing the strategy pool so that the finally selected first strategy can better adapt to the market environment, thereby improving the performance and robustness of the trading strategy.

[0055] In some optional implementations of this embodiment, step S205 includes the following steps: The market status is updated based on the real-time market data to obtain the corresponding target market status.

[0056] In this embodiment, the market state vector can be updated in a timely manner based on newly collected real-time market data to obtain the target market state. Specifically, each time new market data arrives, the various feature values ​​in the market state vector are recalculated to ensure that the market state accurately reflects the current market situation. For example, when new stock price data arrives, technical indicators such as the moving average of stock prices and the relative strength index are recalculated.

[0057] The first strategy and the target market state are calculated based on the preset strategy gradient method to obtain the corresponding strategy gradient.

[0058] In this embodiment, a reinforcement learning-based policy gradient method is employed to calculate the corresponding policy gradient based on the current policy (first policy), the target market state, the trading action, and the reward. The policy gradient represents the direction of policy parameter updates, reflecting the degree to which adjusting the policy parameters affects policy performance under the current target market state. For example, if increasing the value of a certain policy parameter improves the expected return of the strategy, the policy gradient is positive in the direction of that parameter; otherwise, it is negative. Various methods can be used to calculate the policy gradient, such as the Monte Carlo policy gradient method and the temporal difference policy gradient method. The specific method chosen depends on the characteristics of the trading strategy and market demand.

[0059] Preferably, the above-mentioned strategy gradient method is updated online using the following formula:

[0060] The parameters in the formula are explained as follows: Meaning: Objective function (Typically, this refers to the expected cumulative reward of the strategy) Regarding parameters The gradient, its function is to guide how to adjust the policy parameters. To maximize ). Meaning: Expected value, representing the expected value in a strategy The statistical average of the trajectory (state-action sequence) is used to approximate the gradient by sampling multiple trajectories (MonteCarlo estimation) or a single trajectory (such as REINFORCE). Meaning: Policy network in state Select action The probability, its function: to represent the current policy's preference for the action, gradient. Used to update parameters to increase the probability of high-reward actions. Meaning: The gradient of the logarithmic policy probability is called the score function. Its function is to guide the gradient's ascent direction. >0, increase The probability increases; conversely, it decreases. Through the chain rule, the policy parameters are... Associated with reward signals. Meaning: moment The immediate reward, calculated by a predefined reward function, directly reflects the action. The quality of the gradient is the signal source for gradient updates. Meaning: Baseline term, usually an estimate of historical rewards (such as an exponential moving average), function: to reduce variance; if not subtracted... The variance of the gradient estimate may be large (because...) (Vast fluctuations). Does not introduce bias: mathematically... It is independent of the action and does not affect the gradient expectation.

[0061] The physical meaning of the formula is: Core idea: Optimize the strategy by increasing the probability of actions leading to high rewards and decreasing the probability of actions leading to low rewards. The role of the baseline: If... > If the gradient is positive, the policy tends to repeat. .like < When the gradient is negative, the policy tends to avoid... The baseline standardizes the reward signal (similar to batch normalization), reducing update variance. This allows for maximizing the expected reward through gradient ascent while balancing efficiency and stability through the baseline technique.

[0062] The learning rate is adjusted based on the real-time market data to obtain the corresponding target learning rate.

[0063] In this embodiment, the specific implementation process of adjusting the learning rate based on the real-time market data to obtain the corresponding target learning rate will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0064] Based on the policy gradient and the target learning rate, the policy parameters of the first policy are adjusted to obtain the corresponding specified policy.

[0065] In this embodiment, the policy parameters can be updated using the corresponding parameter update formula based on the calculated policy gradient and the dynamically adjusted learning rate. A common parameter update formula is: ,in, Indicates the strategy parameters, Indicates the target learning rate. This represents the policy gradient. By continuously updating the policy parameters, the policy can gradually adapt to market changes, thereby improving its performance.

[0066] Specifically, the policy gradient and target learning rate can be calculated based on the above parameter update formula, and the policy parameters of the first policy can be adjusted according to the calculated updated policy parameters, thereby using the obtained specified policy as the corresponding second policy.

[0067] The specified strategy is used as the second strategy.

[0068] This application updates the market state based on real-time market data to obtain a corresponding target market state. Then, it calculates the first strategy and the target market state using a preset strategy gradient method to obtain a corresponding strategy gradient. Next, it adjusts the learning rate based on the real-time market data to obtain a corresponding target learning rate. Subsequently, it adjusts the strategy parameters of the first strategy based on the strategy gradient and the target learning rate to obtain a corresponding specified strategy. Finally, it uses the specified strategy as the second strategy. Based on this process, this application updates the market state based on real-time market data, and then uses the strategy gradient method to update the first strategy based on the obtained target market state, enabling the trading strategy to be dynamically adjusted according to real-time market data. Furthermore, through strategy gradient calculation and dynamic learning rate adjustment, the strategy parameters can be continuously optimized, allowing the strategy to adapt to changes in market structure in a timely manner, thereby effectively improving the real-time performance and adaptability of the strategy and maintaining good performance in a constantly changing market environment.

[0069] In some optional implementations, the process of adjusting the learning rate based on the real-time market data to obtain the corresponding target learning rate includes the following steps: The real-time market data is analyzed and processed based on preset statistical indicators to obtain corresponding market structure change information.

[0070] In this embodiment, the above-mentioned statistical indicators include at least standard deviation and correlation coefficient.

[0071] Monitoring changes in market structure is fundamental to dynamically adjusting the learning rate. Information about changes in market structure can manifest in various ways, such as increases or decreases in market volatility, shifts in market trends, and sector rotation. Specifically, changes in market structure can be monitored and corresponding information generated by calculating statistical indicators of market data (such as standard deviation and correlation coefficient). For example, if stock price volatility suddenly increases, it indicates that the market has become more unstable, and the learning rate may need to be adjusted.

[0072] Obtain the preset learning rate adjustment rules.

[0073] In this embodiment, the learning rate adjustment rule includes the following: a rule for dynamically adjusting the learning rate (i.e., the learning rate adjustment rule) is pre-defined based on changes in market structure. Specifically, when market volatility is high and changes rapidly, the learning rate should be appropriately reduced to make strategy parameter updates more cautious and avoid over-adjustment of the strategy due to market noise; while when the market is relatively stable and the trend is clear, the learning rate can be appropriately increased to accelerate the convergence speed of the strategy. For example, by setting a volatility threshold, when market volatility exceeds the threshold, the learning rate is reduced to half of its original value; when market volatility is below the threshold, the learning rate is increased to twice its original value.

[0074] Preferably, the adjustment formula corresponding to the above learning rate adjustment rule includes: .in, The adjusted learning rate, The initial learning rate, To adjust the coefficient, This refers to the change in volatility. Volatility The standard deviation of prices over a certain period can be calculated; when volatility increases, it indicates increased market uncertainty. By using a learning rate adjustment rule, the learning rate can be automatically increased to accelerate strategy convergence and enable the strategy to adapt to market changes more quickly.

[0075] Based on the learning rate adjustment rules, the learning rate is adjusted according to the market structure change information to obtain the adjusted specified learning rate.

[0076] In this embodiment, the learning rate can be adjusted based on the rules of the above-mentioned learning rate adjustment rules and the obtained market structure change information, and the adjusted specified learning rate can be used as the corresponding target learning rate.

[0077] The specified learning rate is used as the target learning rate.

[0078] This application analyzes and processes real-time market data based on preset statistical indicators to obtain corresponding market structure change information; then, it obtains preset learning rate adjustment rules; subsequently, based on the learning rate adjustment rules, it adjusts the learning rate according to the market structure change information to obtain an adjusted specified learning rate; finally, it uses the specified learning rate as the target learning rate. Based on the above processing flow, this application analyzes and processes real-time market data using statistical indicators to obtain market structure change information, and then, based on the use of learning rate adjustment rules, it can automatically and intelligently adjust the learning rate according to the market structure change information, and use the adjusted specified learning rate as the corresponding target learning rate. This enables dynamic adjustment of the learning rate, improving the accuracy and adaptability of the generated target learning rate.

[0079] In some alternative implementations, step S207 includes the following steps: Obtain the target weight corresponding to the second strategy.

[0080] In this embodiment, the specific implementation process of obtaining the target weight corresponding to the second strategy will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0081] Call the preset fusion formula.

[0082] In this embodiment, the above fusion formula is specifically as follows: .in, As weight, As a strategy, For action, Given the state, the meaning of this fusion formula is: to obtain the final trading decision by weighting the probability distributions of the actions output by each strategy according to their respective weights.

[0083] The target weight and the output action result are weighted and fused based on the fusion formula to obtain the corresponding calculation data.

[0084] In this embodiment, the target weight and the output action result can be weighted and fused using the above fusion formula. That is, the probability distribution of each second strategy output (output action result) is weighted and averaged according to the updated target weight to obtain the probability calculation data of each action strategy, i.e., the calculation data.

[0085] The calculated data is analyzed and processed based on a preset maximum probability principle in order to determine the corresponding target action strategy from the output action results.

[0086] In this embodiment, each of the aforementioned second strategies (assuming there are two strategies, Strategy 1 and Strategy 2) outputs an action probability distribution (including the probability of selling, the probability of holding, and the probability of buying) based on the current target market state. For example, Strategy 1 outputs a selling probability of 0.3, a holding probability of 0.4, and a buying probability of 0.3; Strategy 2 outputs a selling probability of 0.2, a holding probability of 0.3, and a buying probability of 0.5. The action probability distributions output by each strategy are weighted and averaged to obtain the final trading decision, i.e., the target action strategy. Assuming the weight of Strategy 1 is 0.6 and the weight of Strategy 2 is 0.4, then the final selling probability is 0.6. 0.3 + 0.4 0.2 = 0.26, the probability of holding the position is 0.6. 0.4 + 0.4 0.3 = 0.36, the probability of buying is 0.6. 0.3 + 0.4 0.5 = 0.38. Based on the principle of maximum probability, the final choice is to buy (with the highest probability value) as the trading action, i.e., the target action strategy.

[0087] The target action strategy is used as the target processing strategy.

[0088] This application obtains the target weight corresponding to the second strategy; then calls a preset fusion formula; and performs weighted fusion processing on the target weight and the output action result based on the fusion formula to obtain corresponding calculated data; subsequently, it analyzes and processes the calculated data based on a preset maximum probability principle to determine the corresponding target action strategy from the output action result; and then uses the target action strategy as the target processing strategy. Based on the above processing flow, the dynamic weight fusion and strategy switching processing method provided by this application enables automatic switching to the current optimal strategy while retaining the diversity of suboptimal strategies, avoiding the risk of a single strategy failing. By dynamically adjusting the weights according to the fitness of the strategy, the advantages of multiple strategies are complemented, avoiding the limitations that a single strategy may have. At the same time, the flexible switching of strategies according to market conditions makes the final trading decision (target processing strategy) more reasonable, better adaptable to different market environments, and thus effectively improves the overall trading returns and risk control capabilities.

[0089] In some optional implementations of this embodiment, obtaining the target weight corresponding to the second strategy includes the following steps: Obtain the initial weights corresponding to the second strategy.

[0090] In this embodiment, the initial weights corresponding to the second weights can be generated in advance using an initial weights method. Specifically, the method for initializing weights can be determined based on the initial performance of the strategy. For example, strategies that perform well in the initial test can be given higher initial weights; alternatively, a uniform weight distribution method can be used, where each strategy has the same initial weight to ensure that each strategy has the same influence in the initial stage.

[0091] Calculate the second fitness of the second strategy.

[0092] In this embodiment, the second fitness refers to the fitness score of the second strategy in the current round of trading. The process of calculating the second fitness of the second strategy can be referred to the specific processing procedure described above for calculating the fitness corresponding to each strategy using the historical market data, and will not be elaborated upon here.

[0093] Get the preset soft maximum rule.

[0094] In this embodiment, the soft maximum rule described above is a method for converting fitness scores into weights, which allows strategies with good recent performance to receive higher weights. The specific weight calculation formula is as follows: ,in, Indicates the first The weights of each strategy Indicates the first The fitness of each strategy The sum of the fitness scores for all strategies is represented by an index. This is a temperature coefficient used to control the concentration of weight allocation. When the value is large, weight allocation is more concentrated on strategies with high fitness; when When the weights are relatively small, the weight distribution is relatively even. This allows strategies with good recent performance to receive higher weights, while poorly performing strategies are gradually eliminated. Updating weights using the soft maximum rule allows for a more reasonable weight distribution, highlighting high-performing strategies.

[0095] Based on the second fitness, the initial weights are updated using the soft maximum rule to obtain the corresponding updated weights.

[0096] In this embodiment, the initial weights and the second fitness of the second strategy can be calculated based on the weight calculation formula corresponding to the soft maximum rule, and the updated weights obtained can be used as the target weights of the corresponding second strategy.

[0097] The updated weight is used as the target weight corresponding to the second strategy.

[0098] This application obtains initial weights corresponding to the second strategy; then calculates the second fitness of the second strategy; subsequently obtains a preset soft maximum rule; then, based on the second fitness, uses the soft maximum rule to update the initial weights to obtain the corresponding updated weights; finally, the updated weights are used as the target weights corresponding to the second strategy. Based on the above processing flow, this application obtains initial weights corresponding to the second strategy, calculates the second fitness of the second strategy, and then updates the initial weights based on the calculated second fitness and the use of the soft maximum rule, using the resulting updated weights as the target weights of the corresponding second strategy. Thus, by using the soft maximum rule to update the weights of the second strategy, this application enables more reasonable weight allocation and highlights strategies with excellent performance, effectively improving the accuracy and intelligence of the generated target weights.

[0099] In some optional implementations of this embodiment, after step S207, the electronic device may further perform the following steps: Generate a policy version number corresponding to the target processing policy.

[0100] In this embodiment, rules for generating policy version numbers are pre-defined, and the version numbers should clearly reflect the update order and changes of the policy. Specifically, semantic version number rules can be used, such as major version number.minor version number.patch number (MAJOR.MINOR.PATCH). The major version number indicates a major policy update, the minor version number indicates the addition or improvement of policy functionality, and the patch number indicates the bug fix of the policy. Each time the policy is updated, a corresponding policy version number will be generated according to the type and extent of the update.

[0101] Generate a hash signature corresponding to the target processing strategy.

[0102] In this embodiment, the parameters (i.e., model weights) of the current strategy (i.e., the final generated target processing strategy) can be used. ), the fitness score of the target processing strategy in this round of training or trading (such as return rate or Sharpe ratio), ) and timestamp The data is concatenated and then a hash algorithm (such as Sha256) is used to generate a hash signature. .

[0103] Call the preset blockchain platform.

[0104] In this embodiment, the selection of the aforementioned blockchain platform is not specifically limited; a suitable blockchain platform can be chosen for on-chain recording based on actual business needs. The blockchain platform should possess high availability, security, and scalability, and be able to meet the requirements of policy audit recording. For example, well-known blockchain platforms such as Ethereum and Hyperledger can be selected.

[0105] The strategy version number and the hash signature are recorded on the blockchain platform.

[0106] In this embodiment, the generated strategy version number and hash signature are recorded on the blockchain. Specifically, a transaction is created on the selected blockchain platform, and the strategy version number and hash signature are stored as part of the transaction's data. Simultaneously, a timestamp for recording on the blockchain is recorded to ensure the accuracy of the recorded time.

[0107] Additionally, audit logs for each strategy update can be uploaded to the blockchain in hash form. These logs contain detailed information such as the update time, content, and reason. The audit logs are then hashed, and the hash value is recorded on the blockchain, forming a verifiable audit chain. This approach ensures the traceability, verifiability, and immutability of the strategy evolution process, facilitating regulatory review of historical risk records. For example, regulators can query records on the blockchain to understand the strategy's update history and changes, and assess the strategy's compliance and risk level.

[0108] This application generates a strategy version number corresponding to the target processing strategy; generates a hash signature corresponding to the target processing strategy; then invokes a preset blockchain platform; subsequently, the strategy version number and hash signature are recorded on the blockchain platform. Based on the above processing flow, this application achieves transparency and traceability of the strategy evolution process by generating strategy versions and hash signatures and recording them on the blockchain, enhancing the credibility and compliance of the trading strategy, which is beneficial for supervision and risk control, and ensures the stable operation of the trading system.

[0109] In some alternative implementations, the user information obtained is subject to user consent and complies with relevant laws and policies.

[0110] Furthermore, any software tools or components not belonging to our company that appear in the embodiments of this application are merely illustrative examples and do not represent actual use.

[0111] Furthermore, compared to traditional static quantitative models, this application achieves a leap from "one-off strategy design" to "continuous evolution and adaptive optimization." Reinforcement learning enables the strategy to learn market behavior patterns in real time, while evolutionary algorithms provide global search capabilities, avoiding getting trapped in local optima. The combination of these two allows the system to consistently maintain a high Sharpe ratio and stable returns in volatile markets. Moreover, through a dynamic weighting mechanism and blockchain-based auditing, this system achieves significant improvements in transparency and security. The entire strategy update process is recorded on the blockchain, allowing for compliance and risk control review; while the multi-strategy fusion mechanism prevents the risk of single model failure. In summary, this application truly endows quantitative strategies with "dynamic adaptability + traceability + self-learning capability," representing a key innovative direction for future intelligent trading systems.

[0112] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0113] It should be emphasized that, in order to further ensure the privacy and security of the above target processing strategy, the above target processing strategy can also be stored in a node of a blockchain.

[0114] The blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0115] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results. Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0116] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware through computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0117] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0118] Further reference Figure 3 As a response to the above Figure 2 To implement the method shown, this application provides an embodiment of an artificial intelligence-based data generation device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0119] like Figure 3 As shown, the artificial intelligence-based data generation device 300 described in this embodiment includes: a first processing module 301, a generation module 302, an optimization module 303, a collection module 304, an update module 305, an acquisition module 306, a fusion module 307, and an output module 308. Wherein: The first processing module 301 is used to collect historical market data corresponding to the target product and obtain definition data corresponding to the historical market data; The generation module 302 is used to perform strategy construction processing based on the historical market data and the defined data to generate a corresponding initial strategy pool; The optimization module 303 is used to optimize the initial policy pool based on a preset evolutionary algorithm to select the corresponding first policy; wherein the number of the first policies includes multiple policies. Collection module 304 is used to collect real-time market data corresponding to the target product; The update module 305 is used to update the first strategy based on the real-time market data using a preset reinforcement learning algorithm to obtain the corresponding second strategy. The acquisition module 306 is used to acquire the output action results corresponding to the real-time market data generated based on each of the second strategies; The fusion module 307 is used to perform fusion processing on the output action result based on a preset fusion method to generate a corresponding target processing strategy; The output module 308 is used to output the target processing strategy.

[0120] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data generation method in the aforementioned implementation method, and will not be repeated here.

[0121] In some optional implementations of this embodiment, the optimization module 303 includes: The first calculation submodule is used to calculate the fitness of each strategy in the initial strategy pool using the historical market data. The operation submodule is used to perform selection, crossover, and mutation operations on the policy based on the fitness to obtain a newly generated third policy. The second calculation submodule is used to calculate the first fitness of the third strategy; An integration submodule is used to integrate the strategy and the third strategy to obtain a corresponding strategy set; The sorting submodule is used to sort the strategy set in descending order of the values ​​based on the fitness and the first fitness, and to obtain the corresponding sorting result. The filtering submodule is used to filter out a specified number of fourth strategies from the sorting results and use the fourth strategies as the first strategy.

[0122] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data generation method in the aforementioned implementation method, and will not be repeated here.

[0123] In some optional implementations of this embodiment, the update module 305 includes: The update submodule is used to perform market status update processing based on the real-time market data to obtain the corresponding target market status. The third calculation submodule is used to calculate the first strategy and the target market state based on the preset strategy gradient method to obtain the corresponding strategy gradient. The first adjustment submodule is used to adjust the learning rate based on the real-time market data to obtain the corresponding target learning rate; The second adjustment submodule is used to adjust the policy parameters of the first policy based on the policy gradient and the target learning rate to obtain the corresponding specified policy. The first determining submodule is used to use the specified strategy as the second strategy.

[0124] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data generation method in the aforementioned implementation method, and will not be repeated here.

[0125] In some optional implementations of this embodiment, the first adjustment submodule includes: The analysis unit is used to analyze and process the real-time market data based on preset statistical indicators to obtain corresponding market structure change information. The first acquisition unit is used to acquire the preset learning rate adjustment rules; An adjustment unit is used to adjust the learning rate based on the learning rate adjustment rules and the market structure change information to obtain the adjusted specified learning rate. The first determining unit is used to use the specified learning rate as the target learning rate.

[0126] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data generation method in the aforementioned implementation method, and will not be repeated here.

[0127] In some optional implementations of this embodiment, the fusion module 307 includes: The acquisition submodule is used to acquire the target weight corresponding to the second strategy; Call the submodule to invoke the preset fusion formula; The fusion submodule is used to perform weighted fusion processing on the target weight and the output action result based on the fusion formula to obtain the corresponding calculation data; The analysis submodule is used to analyze and process the calculated data based on a preset maximum probability principle, so as to determine the corresponding target action strategy from the output action results; The second determining submodule is used to use the target action strategy as the target processing strategy.

[0128] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data generation method in the aforementioned implementation method, and will not be repeated here. In some optional implementations of this embodiment, the acquisition submodule includes: The second acquisition unit is used to acquire the initial weights corresponding to the second strategy; A computing unit is used to calculate the second fitness of the second strategy; The third acquisition unit is used to acquire the preset soft maximum rule; An update unit is used to update the initial weights based on the second fitness using the soft maximum rule to obtain the corresponding updated weights. The second determining unit is used to take the updated weight as the target weight corresponding to the second strategy.

[0129] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data generation method in the aforementioned implementation method, and will not be repeated here.

[0130] In some optional implementations of this embodiment, the artificial intelligence-based data generation device further includes: The first generation module is used to generate a strategy version number corresponding to the target processing strategy; The second generation module is used to generate a hash signature corresponding to the target processing strategy; The calling module is used to invoke a pre-defined blockchain platform; The second processing module is used to record the strategy version number and the hash signature on the blockchain platform.

[0131] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data generation method in the aforementioned implementation method, and will not be repeated here. To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0132] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected via a system bus. It should be noted that only the computer device 4 with components 41-43 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0133] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0134] The memory 41 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 41 may be an internal storage unit of the computer device 4, such as the hard disk or memory of the computer device 4. In other embodiments, the memory 41 may also be an external storage device of the computer device 4, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 4. Of course, the memory 41 may include both the internal storage unit and its external storage device of the computer device 4. In this embodiment, the memory 41 is typically used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions for data generation methods based on artificial intelligence. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or will be output.

[0135] In some embodiments, the processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 42 is typically used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to execute computer-readable instructions stored in the memory 41 or to process data, for example, to execute computer-readable instructions of the artificial intelligence-based data generation method.

[0136] The network interface 43 may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the computer device 4 and other electronic devices.

[0137] Compared with the prior art, the embodiments of this application have the following beneficial effects: In this embodiment, the application generates an initial strategy pool by constructing strategies based on collected historical market data and acquired definition data. Then, it optimizes the initial strategy pool using an evolutionary algorithm to select a first strategy. Next, based on collected real-time market data, it updates the first strategy using a reinforcement learning algorithm to obtain a second strategy. It then obtains the output action results corresponding to the real-time market data generated by each second strategy. Finally, it fuses the output action results using a fusion method to generate and output the final target processing strategy. Thus, by first generating an initial strategy pool based on historical market data and optimizing it using an evolutionary algorithm, and then making real-time decisions and updating strategies based on actual market data using a combination of reinforcement learning algorithms and fusion methods, the application achieves intelligent and accurate automatic generation of strategies related to the target product, effectively improving the processing efficiency and flexibility of strategy generation.

[0138] This application also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the artificial intelligence-based data generation method described above.

[0139] Compared with the prior art, the embodiments of this application have the following main advantages: In this embodiment, the application generates an initial strategy pool by constructing strategies based on collected historical market data and acquired definition data. Then, it optimizes the initial strategy pool using an evolutionary algorithm to select a first strategy. Next, based on collected real-time market data, it updates the first strategy using a reinforcement learning algorithm to obtain a second strategy. It then obtains the output action results corresponding to the real-time market data generated by each second strategy. Finally, it fuses the output action results using a fusion method to generate and output the final target processing strategy. Thus, by first generating an initial strategy pool based on historical market data and optimizing it using an evolutionary algorithm, and then making real-time decisions and updating strategies based on actual market data using a combination of reinforcement learning algorithms and fusion methods, the application achieves intelligent and accurate automatic generation of strategies related to the target product, effectively improving the processing efficiency and flexibility of strategy generation.

[0140] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0141] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

Claims

1. A data generation method based on artificial intelligence, characterized in that, Includes the following steps: Collect historical market data corresponding to the target product, and obtain definition data corresponding to the historical market data; Based on the historical market data and the defined data, strategy construction processing is performed to generate a corresponding initial strategy pool; The initial policy pool is optimized based on a preset evolutionary algorithm to select the corresponding first policy; wherein, the number of the first policies includes multiple policies. Collect real-time market data corresponding to the target product; Based on the real-time market data, a preset reinforcement learning algorithm is used to update the first strategy to obtain the corresponding second strategy. Obtain the output action results corresponding to the real-time market data generated based on each of the second strategies; The output action results are fused based on a preset fusion method to generate a corresponding target processing strategy; The target processing strategy is then processed for output.

2. The data generation method based on artificial intelligence according to claim 1, characterized in that, The step of optimizing the initial policy pool based on a preset evolutionary algorithm to select the corresponding first policy specifically includes: For each strategy in the initial strategy pool, the fitness corresponding to each strategy is calculated using the historical market data; Based on the fitness, the policy is subjected to selection, crossover, and mutation operations to obtain a newly generated third policy. Calculate the first fitness of the third strategy; The strategy and the third strategy are integrated to obtain a corresponding strategy set; Based on the fitness and the first fitness, the strategy set is sorted in descending order of value to obtain the corresponding sorting result; A specified number of fourth strategies are selected from the sorting results, and the fourth strategies are used as the first strategy.

3. The data generation method based on artificial intelligence according to claim 1, characterized in that, The step of updating the first strategy using a preset reinforcement learning algorithm based on the real-time market data to obtain the corresponding second strategy specifically includes: Based on the real-time market data, the market status is updated to obtain the corresponding target market status. The first strategy and the target market state are calculated based on the preset strategy gradient method to obtain the corresponding strategy gradient; The learning rate is adjusted based on the real-time market data to obtain the corresponding target learning rate; Based on the policy gradient and the target learning rate, the policy parameters of the first policy are adjusted to obtain the corresponding specified policy; The specified strategy is used as the second strategy.

4. The data generation method based on artificial intelligence according to claim 3, characterized in that, The step of adjusting the learning rate based on the real-time market data to obtain the corresponding target learning rate specifically includes: The real-time market data is analyzed and processed based on preset statistical indicators to obtain corresponding market structure change information. Obtain the preset learning rate adjustment rules; Based on the learning rate adjustment rule, the learning rate is adjusted according to the market structure change information to obtain the adjusted specified learning rate; The specified learning rate is used as the target learning rate.

5. The data generation method based on artificial intelligence according to claim 1, characterized in that, The step of fusing the output action result based on a preset fusion method to generate a corresponding target processing strategy specifically includes: Obtain the target weight corresponding to the second strategy; Call the preset fusion formula; Based on the fusion formula, the target weight and the output action result are weighted and fused to obtain the corresponding calculation data; The calculated data is analyzed and processed based on the preset maximum probability principle in order to determine the corresponding target action strategy from the output action results; The target action strategy is used as the target processing strategy.

6. The data generation method based on artificial intelligence according to claim 5, characterized in that, The step of obtaining the target weight corresponding to the second strategy specifically includes: Obtain the initial weights corresponding to the second strategy; Calculate the second fitness of the second strategy; Retrieve the preset soft maximum rule; Based on the second fitness, the soft maximum rule is used to update the initial weights to obtain the corresponding updated weights; The updated weight is used as the target weight corresponding to the second strategy.

7. The data generation method based on artificial intelligence according to claim 1, characterized in that, After the step of fusing the output action result based on a preset fusion method to generate a corresponding target processing strategy, the method further includes: Generate a policy version number corresponding to the target processing policy; Generate a hash signature corresponding to the target processing strategy; Invoke the preset blockchain platform; The strategy version number and the hash signature are recorded on the blockchain platform.

8. A data generation device based on artificial intelligence, characterized in that, include: The first processing module is used to collect historical market data corresponding to the target product and obtain definition data corresponding to the historical market data; The generation module is used to perform strategy construction processing based on the historical market data and the defined data to generate a corresponding initial strategy pool. An optimization module is used to optimize the initial policy pool based on a preset evolutionary algorithm to select a corresponding first policy; wherein the number of the first policies includes multiple policies. The collection module is used to collect real-time market data corresponding to the target product; The update module is used to update the first strategy based on the real-time market data using a preset reinforcement learning algorithm to obtain the corresponding second strategy. The acquisition module is used to acquire the output action results corresponding to the real-time market data generated based on each of the second strategies; The fusion module is used to perform fusion processing on the output action results based on a preset fusion method to generate a corresponding target processing strategy; The output module is used to perform output processing on the target processing strategy.

9. A computer device, characterized in that, The system includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the data generation method based on artificial intelligence as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the data generation method based on artificial intelligence as described in any one of claims 1 to 7.

Images