Data processing method, apparatus, device, medium, and program product

Abstract

This application discloses a data processing method, apparatus, device, storage medium, and program product. The method includes: selecting a set of stocks that have hit their daily limit up from historical data; filtering the set of stocks that have hit their daily limit up based on market conditions to obtain a first candidate stock set; filtering the first candidate stock set based on technical factors to obtain a second candidate stock set; inputting the second candidate stock set into a network model for intelligent prediction to obtain a list of predicted profit probabilities corresponding to the second candidate stock set output by the network model; and selecting a target stock set from the second candidate stock set based on the list of predicted profit probabilities.

Description

Technical Field

[0001] This application relates to the field of financial transaction technology, and in particular to a data processing method, apparatus, device, storage medium, and program product. Background Technology

[0002] In the field of securities investment, investors are always striving to find analytical methods that can effectively predict stock prices, especially capturing opportunities for sharp short-term price increases (such as limit-up moves). Traditional stock selection methods mainly rely on two categories of analysis: fundamental analysis and technical analysis. Fundamental analysis focuses on assessing long-term factors such as a company's financial statements, industry prospects, and the macroeconomy; while technical analysis mainly relies on historical stock prices and trading volume data, using chart patterns and technical indicators (such as moving averages) to predict future trends. However, both methods rely heavily on the analyst's subjective experience, making it difficult to handle massive amounts of high-speed, multi-dimensional financial data, and their reaction speed and decision-making efficiency are significantly limited in volatile trading markets. Summary of the Invention

[0003] To address the aforementioned technical problems, embodiments of this application provide a data processing method, apparatus, device, storage medium, and program product.

[0004] The data processing method provided in this application includes: Select a set of stocks that hit the daily limit up from historical data; The market status of the set of stocks that hit the daily limit is filtered to obtain the first candidate stock set; The first candidate stock set is screened using technical factors to obtain the second candidate stock set; The second candidate stock set is input into the network model for intelligent prediction, and the predicted profit probability list corresponding to the second candidate stock set output by the network model is obtained. The target stock set is selected from the second candidate stock set based on the predicted profit probability list.

[0005] The data processing apparatus provided in this application embodiment includes: A filtering unit is used to filter out a set of stocks that have hit their daily limit from historical data; to filter the set of stocks that have hit their daily limit based on market status to obtain a first candidate stock set; and to filter the first candidate stock set based on technical factors to obtain a second candidate stock set. The prediction unit is used to input the second candidate stock set into the network model for intelligent prediction and obtain the predicted profit probability list corresponding to the second candidate stock set output by the network model. The filtering unit is further configured to filter out a target stock set from the second candidate stock set based on the predicted profit probability list.

[0006] The processing device provided in this application includes a processor and a memory, wherein the memory is used to store computer programs, and the processor is used to call and run the computer programs stored in the memory to execute any of the above-described data processing methods.

[0007] The computer-readable storage medium provided in this application embodiment is used to store a computer program that causes a computer to execute any of the above-described data processing methods.

[0008] The computer program product provided in this application includes computer program instructions that cause a computer to execute any of the above-described data processing methods.

[0009] In the technical solution of this application embodiment, a set of stocks that have hit their daily limit is selected from historical data; the set of stocks that have hit their daily limit is filtered by market status to obtain a first candidate stock set; the first candidate stock set is filtered by technical factors to obtain a second candidate stock set; the second candidate stock set is input into a network model for intelligent prediction to obtain a list of predicted profit probabilities corresponding to the second candidate stock set output by the network model; and a target stock set is selected from the second candidate stock set based on the list of predicted profit probabilities. Thus, by selecting a set of stocks that have hit their daily limit from historical data and dynamically identifying effective market conditions within that set to obtain a first candidate stock set, we can ensure that the stock selection strategy is only activated under historically validated market conditions, proactively avoiding ineffective trading environments. Subsequently, by using effective technical factors to filter the first candidate stock set to obtain a second candidate stock set, we can identify the characteristics of stocks with subsequent potential for hitting their daily limit. This provides cleaner and higher-quality training data and prediction targets for subsequent intelligent predictions, allowing the network model to focus more on learning high-value patterns. Then, by feeding the second candidate stock set into the network model for intelligent prediction, we can achieve a deep intelligent assessment of stock potential, ensuring the accuracy and directionality of the prediction signals. Finally, based on the predicted profit probability list output by the network model, we select the optimal stock selection results (target stock set), thereby significantly improving the accuracy of stock selection for stocks hitting their daily limit and the stability of investment returns. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating the data processing method provided in an embodiment of this application; Figure 2 This is a flowchart illustrating an AI multi-factor limit-up stock selection analysis method that combines market state recognition and sample selection, as provided in an embodiment of this application. Figure 3 This is a bar chart illustrating the average return of N-day holdings of stocks that have naturally hit their daily price limit under four market conditions, as provided in the embodiments of this application. Figure 4 This is a histogram of the average return of the Williams %R indicator over multiple periods over N days for stocks that have hit their daily price limit. Figure 5 This is a histogram of the average return of the momentum indicator over multiple periods over N days for stocks that have hit their daily price limit. Figure 6 This is a bar chart illustrating the average return on the turnover rate of stocks that have hit their daily price limit for N days under the natural limit-up holding period, provided in this application embodiment. Figure 7 This is a bar chart illustrating the average return on the number of daily limit-up days under N days of holding naturally limit-up stocks, provided in this application embodiment. Figure 8 This is a schematic diagram of the net return curve for the next N trading days predicted by the GBT model provided in this application embodiment; Figure 9 This is a schematic diagram of the structure of the data processing apparatus provided in the embodiments of this application; Figure 10 This is a schematic diagram of the processing device provided in the embodiments of this application. Detailed Implementation

[0011] The technical solutions of the embodiments of this application will now be described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0012] To facilitate understanding of the technical solutions of the embodiments of this application, the relevant technologies of the embodiments of this application are described below. The following relevant technologies are optional solutions and can be combined with the technical solutions of the embodiments of this application in any way, and they all fall within the protection scope of the embodiments of this application.

[0013] With the development of artificial intelligence technology, especially the widespread application of machine learning and deep learning algorithms, quantitative stock selection has become an important tool for modern investment decision-making. Among related technologies, numerous stock selection methods based on AI multi-factor models have emerged. These methods typically extract hundreds or even thousands of feature factors (including but not limited to valuation, growth, momentum, sentiment, and risk factors) that may be related to stock price changes from massive historical data, and use algorithms such as regression, classification, and ensemble learning to build predictive models to identify stocks with a high probability of rising in the future.

[0014] Among the technical fields covered by this application, the most representative related technology is a static AI multi-factor quantitative stock selection method that does not distinguish between micro-market conditions and is trained based on a full set of samples. This method uses a fixed AI model trained on a full set of historical samples to handle stock selection tasks under all market conditions, and assumes that all historical limit-up samples have homogeneous learning value, and that the factors driving limit-up are universal and stable.

[0015] However, the above methods still have technical flaws, mainly stemming from the indiscriminate handling of market micro-states. Specific flaws and their causes include: (1) The model performs poorly in specific market environments because it cannot identify and avoid ineffective stock selection scenarios, resulting in low overall win rate and stability.

[0016] The root cause of this deficiency lies in the lack of a clear method for classifying market conditions in the relevant technical solutions, because the effectiveness of the stop-limit-up stock selection strategy is highly dependent on the current micro-market conditions. For example, when the market index shows a "double bearish contrarian combination" in both the daily K-line and 60-minute K-line technical indicators, or a "combination of a bullish trend in the daily K-line technical indicators and a bearish trend in the 60-minute K-line technical indicators," the overall market faces significant downward pressure, and the sustainability of limit-up stocks is extremely poor. Under such circumstances, the success rate of limit-up stock selection is very low.

[0017] However, static models in related technologies cannot perceive these states and will issue stock selection signals equally in all market conditions. This leads to a large number of invalid or even loss-making trades in unsuitable environments such as "counter-trend" and "pullback", thereby dragging down the overall performance of the model.

[0018] (2) The model learned a large number of noisy samples from inefficient market environments, which reduced its ability to learn high-quality limit-up patterns.

[0019] This flaw stems directly from its full-sample training mode and its neglect of market conditions. Historical data contains a large number of limit-up samples generated under different market conditions. Among them, limit-up data generated in environments such as "counter-trend" are often noisy and do not have the potential for further upward movement.

[0020] However, static models in related technologies learn these samples indiscriminately, attempting to find a universal rule that covers all market environments (including inefficient ones). This forces the model's learning objectives to become vague and contradictory, diluting its focus on learning the more valuable patterns of price surges that emerge in highly efficient environments such as "trend-following combinations" and "rebound combinations." Ultimately, this results in the model issuing signals that are not accurate enough even in its efficient market environments.

[0021] In summary, the fundamental flaw of the relevant technologies lies in the fact that their models are state-independent. They cannot focus on high-value state samples during training, nor can they enable or disable strategies based on market conditions during application. This design is completely disconnected from the micro-state of the market, which is the core reason for the unstable performance of the models in the relevant technologies.

[0022] Based on the above analysis of the relevant technologies and their shortcomings, this application aims to solve the following technical problems: (1) The timing failure problem caused by the model's inability to identify the micro-state of the market.

[0023] Static models in related technologies cannot perceive different market states (such as bullish continuation, bearish continuation, pullback, and rebound), causing them to still generate a large number of stock selection signals in market environments unsuitable for chasing limit-up stocks, such as "bearish continuation" or "pullback," resulting in an extremely high failure rate and unnecessary drawdowns. Therefore, this application requires designing a method that can automatically and accurately identify the current market micro-state and ensure that the stock selection strategy is activated only in efficient market states (such as trend following and rebound) that match the limit-up pattern, thereby fundamentally avoiding ineffective operations and improving the overall win rate and stability.

[0024] (2) The problem of insufficient purity of model signal and directional ambiguity caused by training with full sample.

[0025] Existing technologies perform indiscriminate learning on all limit-up samples, forcing the model to fit a large number of noisy limit-up samples from inefficient market environments with low subsequent upside potential. This severely dilutes the model's ability to learn the patterns of high-value limit-up stocks. Therefore, this application requires pre-screening high-quality samples from historical data before model training. These samples not only exhibit the form of limit-up but, more importantly, possess sustained upside potential in the short term after the limit-up. This improves the purity of the training dataset, guiding the AI ​​model to focus more on learning the characteristics of limit-up stocks with genuine profit potential, ultimately achieving more precise directionality in the model's output signals.

[0026] To address the aforementioned technical issues, this application proposes a data processing method, specifically an AI-powered multi-factor limit-up stock selection method that combines market state identification and sample selection. First, high-quality limit-up samples are selected from historical data. Then, the effective market states within the limit-up samples are dynamically identified based on the multi-period Moving Average Convergence Divergence (MACD) indicator. Subsequently, effective technical factors are used for preliminary screening of candidate stocks, which are then fed into an AI multi-factor model for intelligent prediction. Finally, the optimal stock selection result is output based on the model's comprehensive score. This method ensures that the stock selection strategy is only activated in suitable market conditions, targeting stocks with high potential, thereby significantly improving the accuracy of limit-up stock selection and the stability of investment returns.

[0027] To facilitate understanding of the technical solutions of the embodiments of this application, the technical solutions of this application are described in detail below through specific embodiments. The above-mentioned related technologies are optional solutions and can be arbitrarily combined with the technical solutions of the embodiments of this application, all of which fall within the protection scope of the embodiments of this application. The embodiments of this application include at least some of the following contents.

[0028] This application provides a data processing method. Figure 1 This is a flowchart illustrating the data processing method provided in an embodiment of this application, as shown below. Figure 1 As shown, the method includes the following steps: Step 101: Filter out the set of stocks that hit the daily limit from historical data.

[0029] In this embodiment, historical data refers to the K-line chart data of all stocks in the market. Specifically, historical data includes trading data for multiple stocks over multiple first trading days, with each first trading day constituting a historical time period. The trading data includes core price information such as the opening price, closing price, highest price, and lowest price. By analyzing the trading data of multiple stocks over multiple first trading days, it is possible to determine which stocks on each first trading day reached the exchange's daily price limit (limit-up price), triggering the price restriction mechanism. This identifies one or more limit-up stocks on each first trading day, allowing for the selection of a set of limit-up stocks from historical data.

[0030] In some embodiments, step 101 includes: For each of the multiple first trading days, obtain the closing price and opening price of each stock in the multiple stocks on each first trading day. If there are one or more stocks that have reached the daily limit price on each first trading day and whose closing price is equal to the daily limit price and whose opening price is lower than the closing price, then select one or more stocks that have reached the daily limit price from the multiple stocks on each first trading day, and determine the set of stocks that have reached the daily limit price based on one or more stocks that have reached the daily limit price.

[0031] Here, the opening price refers to the price of the first transaction after the start of a single trading day, the closing price refers to the price of the last transaction before the end of a single trading day, and the upper limit price refers to the highest price limit allowed to rise within a single trading day, which aims to curb drastic price fluctuations, prevent excessive speculation, and maintain market stability.

[0032] Specifically, for each first trading day, the opening and closing prices of each stock on that first trading day are obtained. If there are one or more stocks on that first trading day whose closing price is equal to the limit-up price and whose opening price is lower than the closing price, then these one or more stocks are identified as one or more limit-up stocks. The above judgment criteria can ensure that the limit-up is the result of intraday trading, rather than being unable to be bought because the stock opens at the limit-up price. In this way, one or more limit-up stocks on each first trading day can be identified. Based on one or more limit-up stocks on each first trading day, a set of limit-up stocks can be formed.

[0033] Step 102: Filter the stocks that have hit the daily limit up based on market conditions to obtain the first candidate stock set.

[0034] In this embodiment, the multi-period MACD indicator of the market index of each stock in the limit-up stock set is calculated, and the market state of each stock is determined based on the multi-period MACD indicator of the market index of each stock. If the market state of a stock indicates that the stock is trading or investing in line with the current main trend of the market (trend-following market state), then the stock is determined as the first candidate stock. Alternatively, if the market state of a stock indicates that the asset price of the stock has rebounded in a short period of time in a downtrend (rebound market state), then the stock is also determined as the first candidate stock. In this way, the first candidate stock set can be selected from the limit-up stock set based on the market state of each stock.

[0035] In some embodiments, step 102 includes: Get the market index of each stock that hits its daily limit in the set of stocks that hit their daily limit. The market index of each stock that hit the daily limit was calculated in the first and second periods respectively, and the first smoothed moving average of the difference MACD indicator and the second MACD indicator corresponding to each stock that hit the daily limit were obtained. For each stock that hits its daily limit up, if both the first MACD indicator and the second MACD indicator corresponding to that stock are in the bullish zone, then that stock is designated as the first candidate stock; or... If the first MACD indicator corresponding to the stock that hits its daily limit is in the downtrend and the second MACD indicator corresponding to the stock that hits its daily limit is in the uptrend, then the stock that hits its daily limit is identified as the first candidate stock.

[0036] Here, the MACD indicator is a trend-following technical analysis tool used in financial markets, primarily for determining the direction, strength, and potential reversal signals of stock price trends. The MACD indicator mainly includes the Difference (DIF) line, the Differential Exponential Average (DEA) line, and the MACD histogram. The DIF line reflects the degree of deviation between short-term and long-term trends. The DEA line smooths out fluctuations in the DIF line and reduces noise.

[0037] Specifically, firstly, obtain the market index (such as the CSI All Share Index) for each stock that hits its daily limit, and then calculate the MACD indicator for the market index of each stock on the first period (such as one day) and the second period (such as 60 minutes) to obtain the first MACD indicator and the second MACD indicator corresponding to each stock that hits its daily limit; then, based on the first MACD indicator and the second MACD indicator corresponding to each stock that hits its daily limit, determine the market state of each stock that hits its daily limit.

[0038] If there are one or more stocks that have hit their daily limit and both the first and second MACD indicators are in the bullish zone (the DIF line and DEA line are both above the 0 axis, or the MACD histogram (red line) continues to expand or remain positive above the 0 axis), it indicates that these stocks are in a strong upward trend and the bulls are in control. These stocks can be identified as one or more first candidate stocks.

[0039] Alternatively, if there are one or more stocks that have hit their daily limit up and whose first MACD indicator is in the bearish zone (both the DIF and DEA lines are below the zero axis, or the DIF line is below the DEA line) and whose second MACD indicator is in the bullish zone, then it indicates that although these stocks are in a daily downtrend, they have shown short-term oversold rebound momentum. These stocks can also be identified as one or more first-candidate stocks. The one or more first-candidate stocks identified above can collectively form a first-candidate stock set.

[0040] Step 103: Perform technical factor screening on the first candidate stock set to obtain the second candidate stock set.

[0041] In this application embodiment, a set of effective technical factors, such as multi-period Williams %R, multi-period trend amplitude, daily turnover rate on limit-up days, and cumulative number of times the limit-up is touched during trading, can be used to screen the first candidate stock set in order to further screen out limit-up stocks with subsequent potential and obtain the second candidate stock set.

[0042] In some embodiments, step 103 includes: Obtain the preset set of technical factors; The first candidate stock set is screened using technical factors to obtain the second candidate stock set.

[0043] Here, the set of technical factors includes one or more of the following: multi-period Williams %R, multi-period trend amplitude, turnover rate on limit-up days, and cumulative number of times the price hits the limit-up price during the trading session.

[0044] The Williams %R indicator is mainly used to measure overbought and oversold conditions. Multi-period Williams %R intervals refer to whether the stock price is simultaneously in an extremely oversold or oversold zone at different time periods (such as 15 minutes, 60 minutes, and daily charts).

[0045] Multi-period trend amplitude refers to the strength of stock price trends across different time periods, such as the magnitude of the increase, the slope, or the divergence of moving averages.

[0046] The turnover rate on the day a stock hits its daily limit is the ratio of its trading volume to its circulating shares on that day, and is used to measure the quality of a stock hitting its daily limit.

[0047] The cumulative number of times a stock hits its daily price limit refers to the total number of times the stock price touches the daily price limit during the intraday price movement within a single trading day.

[0048] In some embodiments, a second candidate stock set is obtained by screening the first candidate stock set based on a set of technical factors, including one of the following: For each first candidate stock in the first candidate stock set, indicators are calculated for each first candidate stock over multiple periods to obtain multiple Williams indicators for each first candidate stock; the multiple Williams indicators for each first candidate stock are summed to obtain the total Williams indicator for each first candidate stock; if there are one or more first candidate stocks with a total Williams indicator less than a first threshold, then one or more first candidate stocks are determined as one or more second candidate stocks. For each first candidate stock, the indicators are calculated over multiple periods to obtain multiple momentum indicators corresponding to each first candidate stock; the multiple momentum indicators corresponding to each first candidate stock are summed to obtain the total momentum indicator corresponding to each first candidate stock; if there are one or more first candidate stocks with a total momentum indicator less than a second threshold, then one or more first candidate stocks are determined as one or more second candidate stocks. Obtain the turnover rate of each first candidate stock on the first trading day; if there are one or more first candidate stocks with a turnover rate less than the third threshold, then determine one or more first candidate stocks as one or more second candidate stocks. Obtain the number of limit-up days for each first candidate stock on the first trading day. If there are one or more first candidate stocks with a number of limit-up days greater than the fourth threshold, then determine one or more first candidate stocks as one or more second candidate stocks.

[0049] Specifically, in the process of screening the first candidate stock set based on the set of technical factors, any of the following screening methods can be used, and when any of the following screening conditions are met, one or more second candidate stocks are obtained and a second candidate stock set is formed: The first screening method involves calculating the Williams %R for each first-candidate stock across multiple periods (e.g., 10, 25, 50, 90, and 200 days), resulting in multiple Williams %R values ​​for each first-candidate stock. These multiple Williams %R values ​​are then summed to obtain the total Williams %R value for each first-candidate stock. For each first-candidate stock, if the total Williams %R value is less than a first threshold (e.g., 90), it indicates that the first-candidate stock is oversold and has a rebound potential. This first-candidate stock can then be removed from the first-candidate stock set and designated as a second-candidate stock.

[0050] The second screening method involves calculating the momentum index for each first candidate stock across multiple periods (e.g., 10, 25, 50, 90, and 200 days), resulting in multiple momentum indices for each first candidate stock. These multiple momentum indices are then summed to obtain the total momentum index for each first candidate stock. For each first candidate stock, if the total Williams %R index is less than a second threshold (e.g., -0.55), it indicates that although the first candidate stock may experience a short-term decline, its momentum may be exhausted, and this first candidate stock can be removed from the first candidate stock set and used as a second candidate stock.

[0051] The third screening method involves obtaining the turnover rate of each first-candidate stock on its limit-up day (the corresponding first trading day). For each first-candidate stock, if its turnover rate on the limit-up day is less than the third threshold (e.g., 2%), it indicates that the limit-up of the first-candidate stock has not excessively consumed market funds, and there may still be room for further growth. This first-candidate stock can be selected from the first-candidate stock set and used as the second-candidate stock.

[0052] The fourth screening method involves obtaining the number of times each first candidate stock hits its daily limit up on its first trading day (the corresponding first trading day). For each first candidate stock, if the number of times it hits its daily limit up on its first trading day is greater than the fourth threshold (e.g., 10 times), it indicates that the first candidate stock has a strong intention to hit its daily limit up and the quality of its limit-up is high. This first candidate stock can be selected from the first candidate stock set and used as the second candidate stock.

[0053] Step 104: Input the second candidate stock set into the network model for intelligent prediction, and obtain the predicted profit probability list corresponding to the second candidate stock set output by the network model.

[0054] In this embodiment, the network model is specifically an AI multi-factor model, such as a Gradient Boosting Decision Tree (GBDT) model, used to predict the profitability probability of a stock based on its set of feature factors. By inputting the set of second candidate stocks into the AI ​​multi-factor model for intelligent prediction, the AI ​​multi-factor model can analyze and intelligently predict the feature factor set of each second candidate stock in the set, outputting the predicted profitability probability for each second candidate stock, thus obtaining a list of predicted profitability probabilities for the set of second candidate stocks. This list of predicted profitability probabilities includes the price change (label) of each second candidate stock over multiple future trading days (multiple second trading days) and its corresponding profitability probability.

[0055] In some embodiments, the training process of the network model includes: Based on the first MACD indicator and the second MACD indicator corresponding to each second candidate stock in the second candidate stock set, the market technical factor corresponding to each second candidate stock is determined. The overall score for each second candidate stock is determined based on the total Williams %R, total momentum %R, and the number of limit-up days on the first trading day. Based on one or more of the following factors for each second candidate stock: first MACD indicator, second MACD indicator, total Williams %R indicator, total momentum indicator, turnover rate on the first trading day, number of limit-up days on the first trading day, market technical factors, and comprehensive score, determine the set of characteristic factors corresponding to each second candidate stock. The feature factor set corresponding to each second candidate stock is input into the network model for intelligent prediction, and the predicted profit probability list output by the network model is obtained.

[0056] Specifically, firstly, based on the DIF and DEA values ​​of the first and second MACD indicators for each second candidate stock, and considering four stock market states—trend following, rebound, counter-trend, and pullback—the market state for each second candidate stock is determined. Then, based on the market state, the corresponding market technical factors for each second candidate stock are determined. These market technical factors include the corresponding values ​​for the four stock market states: trend following, rebound, counter-trend, and pullback. For example, if both the DIF and DEA values ​​of the first and second MACD indicators for a second candidate stock are above the zero line, then the second candidate stock is determined to be in a trend following market state, and the corresponding value for this state is set to 1; the values ​​for the other market states are set to 0. Finally, a normalization function is used to normalize the total Williams %R and total momentum indicators for each second candidate stock, as well as the number of times each second candidate stock hits its daily limit up, to obtain a comprehensive score for each second candidate stock.

[0057] Next, based on one or more of the following indicators for each second candidate stock: the first MACD indicator, the second MACD indicator, the total Williams %R indicator, the total momentum indicator, the turnover rate and number of limit-up days for each second candidate stock on its limit-up day (first trading day), and the market technical factors and comprehensive scores for each second candidate stock, the set of characteristic factors corresponding to each second candidate stock is determined.

[0058] Then, the feature factor set corresponding to each second candidate stock is input into the network model (AI multi-factor model) for intelligent prediction. The network model performs analysis and intelligent prediction on the feature factor set of each second candidate stock, thereby outputting a list of predicted profit probabilities corresponding to the second candidate stock set.

[0059] Step 105: Select the target stock set from the second candidate stock set based on the predicted profitability probability list.

[0060] In this embodiment, the predicted profitability probability list includes the price change label and profitability probability of each second candidate stock on multiple second trading days. By analyzing the price change label and profitability probability of each second candidate stock in the predicted profitability probability list on each second trading day, one or more optimal stocks with the highest profitability probability in the second candidate stock set can be comprehensively determined, and a target stock set can be formed based on one or more optimal stocks.

[0061] In some embodiments, step 105 includes: Based on the price change label and profit probability of each second candidate stock in multiple second trading days, the set of second candidate stocks is sorted in descending order to obtain the sorted set of second candidate stocks. Select a preset number of second candidate stocks from the sorted second candidate stock set, and determine the target stock set based on the preset number of second candidate stocks.

[0062] Specifically, firstly, the price change (label) and corresponding profit probability of each second candidate stock on each second trading day are comprehensively evaluated to obtain the average profit probability of each second candidate stock over multiple second trading days. The higher the probability, the better the prospects of the corresponding stock. Next, based on the average profit probability of each second candidate stock over multiple second trading days, all second candidate stocks in the second candidate stock set are sorted in descending order to obtain the sorted second candidate stock set. Then, a predetermined number of second candidate stocks are selected from the sorted second candidate stock set. Since the average profit probability of the predetermined number of second candidate stocks over multiple second trading days is optimal, the predetermined number of second candidate stocks can be used as the target stock set for subsequent investors to refer to in stock selection decisions.

[0063] In the technical solution of this application embodiment, a set of stocks that have hit their daily limit is selected from historical data; the set of stocks that have hit their daily limit is filtered by market status to obtain a first candidate stock set; the first candidate stock set is filtered by technical factors to obtain a second candidate stock set; the second candidate stock set is input into a network model for intelligent prediction to obtain a list of predicted profit probabilities corresponding to the second candidate stock set output by the network model; and a target stock set is selected from the second candidate stock set based on the list of predicted profit probabilities. Thus, by selecting a set of stocks that have hit their daily limit from historical data and dynamically identifying effective market conditions within that set to obtain a first candidate stock set, we can ensure that the stock selection strategy is only activated under historically validated market conditions, proactively avoiding ineffective trading environments. Subsequently, by using effective technical factors to filter the first candidate stock set to obtain a second candidate stock set, we can identify the characteristics of stocks with subsequent potential for hitting their daily limit. This provides cleaner and higher-quality training data and prediction targets for subsequent intelligent predictions, allowing the network model to focus more on learning high-value patterns. Then, by feeding the second candidate stock set into the network model for intelligent prediction, we can achieve a deep intelligent assessment of stock potential, ensuring the accuracy and directionality of the prediction signals. Finally, based on the predicted profit probability list output by the network model, we select the optimal stock selection results (target stock set), thereby significantly improving the accuracy of stock selection for stocks hitting their daily limit and the stability of investment returns.

[0064] This application also proposes an AI multi-factor stock selection analysis method that combines market state identification and sample selection. The core idea is to first select high-quality limit-up samples from historical data through a progressively refined processing flow. Then, based on the multi-period MACD indicator, it dynamically identifies the effective market state within these samples. Subsequently, it uses effective technical factors to initially screen candidate stocks, and then inputs the data into an AI multi-factor model for intelligent prediction. Finally, it outputs the optimal stock selection result based on the model's comprehensive score. This method ensures that the stock selection strategy is only activated in suitable market conditions for stocks with high potential, thereby significantly improving the accuracy of limit-up stock selection and the stability of investment returns.

[0065] Figure 2 This is a flowchart illustrating an AI multi-factor limit-up stock selection analysis method that combines market state recognition and sample selection, as provided in an embodiment of this application. Figure 2 As shown, the method includes the following steps: Step 201: Select a set of stocks that hit the daily limit up from historical data and meet specific patterns.

[0066] This step aims to build a high-quality initial sample pool. The system filters out a set of stocks that hit their daily limit up based on specific patterns from historical data. The specific screening criteria are as follows: A stock is considered a limit-up stock if its closing price equals the limit-up price and its opening price is lower than its closing price. This standard ensures that the limit-up is driven by intraday trading, rather than being impossible to buy due to the stock opening at the limit-up price, thus guaranteeing the tradability of the sample and its practical research value.

[0067] Step 202: Determine whether each stock in the limit-up stock set is in a suitable market environment for limit-up stock selection, and select an initial candidate stock set.

[0068] This step is the decision-making center of the entire process, used to determine whether each stock that hits its daily limit is in a suitable market environment for stock selection based on its price limit. Its specific implementation process is as follows: (1) Select the CSI All Share Index (code: 000985) of each stock that hit the daily limit as a representative of the overall market environment.

[0069] (2) Calculate the daily MACD indicator and the 60-minute MACD indicator of the CSI All Share Index for each stock that hit the daily limit, and define the market status based on their direction combination.

[0070] This application found that the limit-up stock selection strategy is most effective in the following two market conditions, therefore the subsequent stock selection process should only be initiated under either of these two market conditions: (1) Follow the trend. The daily MACD indicator is in the bullish zone (golden cross or DIFF>0), and the 60-minute MACD indicator is also in the bullish zone. This indicates that the market is in a strong upward trend.

[0071] (2) Rebound. The daily MACD indicator is in the downtrend (death cross or DIFF<0), while the 60-minute MACD indicator is in the uptrend. This indicates that although the market is in a daily downtrend, there is short-term oversold rebound momentum.

[0072] Step 203: Use a set of effective technical factors to screen the initial candidate stock set to obtain the candidate stock set.

[0073] This step utilizes a set of effective technical factors to quickly screen an initial set of candidate stocks suitable for the market environment, thereby narrowing the scope of the AI ​​model's analysis and improving efficiency. Specific screening criteria (meeting any one of them is sufficient) include: (1) Williams %R percentiles across multiple periods. Calculate the Williams %R for each initial candidate stock across multiple periods, such as 10, 25, 50, 90, and 200 days, and sum the Williams %R values ​​across multiple periods to obtain the total Williams %R for each initial candidate stock. Based on the total Williams %R of each initial candidate stock, select one or more initial candidate stocks from the initial candidate stock set whose total Williams %R is less than 90, indicating that these initial candidate stocks are in an oversold state and have a rebound demand.

[0074] (2) Multi-period trend amplitude. Calculate the momentum index of each initial candidate stock in multiple periods such as 10, 25, 50, 90, and 200 days, and sum the momentum indices of multiple periods to obtain the total momentum index of each initial candidate stock. Based on the total momentum index of each initial candidate stock, select one or more initial candidate stocks from the initial candidate stock set whose total momentum index is less than -0.55, indicating that although these initial candidate stocks may decline in the short term, their momentum may be exhausted.

[0075] (3) Turnover rate on the day of limit up. Select one or more initial candidate stocks from the initial candidate stock set whose turnover rate is less than 2% on the day of limit up. This indicates that the limit up of these initial candidate stocks has not excessively consumed market funds, and there may still be room for further growth.

[0076] (4) The cumulative number of times the stock hits the daily limit during trading. Select one or more initial candidate stocks from the initial candidate stock set that have hit the daily limit more than 10 times in total. This indicates that these initial candidate stocks have a strong desire to hit the daily limit and the quality of the limit-up is high.

[0077] Step 204: Input the candidate stock set into the pre-trained AI model for accurate prediction and obtain a list of predicted profit probabilities corresponding to the candidate stock set.

[0078] This step is the core intelligent decision-making process of the model. The candidate stock set selected in step 203 is input into the pre-trained AI model for accurate prediction, so as to obtain the list of predicted profit probabilities for the candidate stock set output by the AI ​​model.

[0079] The AI ​​model is a classification model built using the gradient boosting decision tree algorithm. Its specific training and prediction process includes: (1) Determine the set of characteristic factors for each candidate stock in the candidate stock set. The set of characteristic factors includes, but is not limited to: the number of times the stock hits the daily limit up, the Williams %R indicator, the momentum indicator, the turnover rate, the DIFF and DEA values ​​of the daily MACD indicator, the DIFF and DEA values ​​of the 60-minute MACD indicator, market technical factors, and a comprehensive score calculated by normalizing the Williams %R indicator, the momentum indicator, and the turnover rate using the MinMaxScaler function. Among them, the market technical factors are discrete values ​​obtained by enumerating four market combination forms obtained from the daily MACD indicator and the 60-minute MACD indicator, including four discrete values: trend following, counter-trend, rebound, and pullback.

[0080] (2) Set training objectives. The model is trained to predict whether a stock will rise in the next N trading days. Specifically, samples that actually rise in the next N trading days are marked as 1 (profit), and samples that fall are marked as 0 (loss).

[0081] (3) The model is iteratively trained based on the training objective to obtain the AI ​​model.

[0082] (4) Input the set of feature factors of each candidate stock into the model for intelligent prediction, and obtain the list of predicted profit probabilities for the candidate stock set output by the AI ​​model.

[0083] Step 205: Based on the predicted profit probability list, select one or more optimal candidate stocks from the candidate stock set as the target stock set.

[0084] This step is the final output of the entire process, aiming to select the most valuable target stocks from the candidate stock set based on the predicted profitability probability list output by the AI ​​model. Its specific implementation process includes: (1) Based on the rise / fall label and profit probability of each candidate stock recorded in the predicted profit probability list in the next N trading days, sort each candidate stock in the candidate stock set in descending order to obtain the sorted candidate stock set. This sorting result comprehensively reflects the relative strength of each candidate stock in multiple key technical indicators.

[0085] (2) Select the top 5 best candidate stocks from the sorted candidate stock set as the target stock set. The target stock set is the final stock selection result output of this application for investors to make decision-making reference.

[0086] This application introduces a layered and collaborative refined processing workflow, organically integrating market state identification, sample selection, and AI prediction to form a highly adaptive technical solution for the specific scenario of selecting stocks hitting their daily limit. This effectively solves the two major shortcomings of related technologies: model state independence and coarse-grained sample selection. Its beneficial effects are specifically reflected in the following four aspects, all of which are directly related to the core steps and modules designed in this application: (1) By screening market conditions, the strategy achieves precise timing, fundamentally avoids ineffective trading environments, and improves the overall win rate and stability.

[0087] Specifically, by combining the daily MACD indicator and the 60-minute MACD indicator of the stock index—two indicators from different timeframes—and through specific combinations of their bullish and bearish directions (resulting in four market states: rebound, pullback, counter-trend, and trend-following), the effective market environment applicable to the limit-up stock selection strategy is precisely defined. Table 1 below shows the average return and daily win rate of naturally limit-up stocks held for N days under the four market states. Figure 3 The diagram shows a bar chart illustrating the average return of naturally limit-up stocks held for N days under four market conditions provided in this application embodiment. The data clearly shows that in the "trend following" and "rebound" market conditions, the subsequent average return and daily win rate of naturally limit-up stocks are significantly higher than in the "pullback" and "counter-trend" market conditions. It is evident that by using market condition screening as the decision switch for the entire stock selection process, subsequent analysis can be ensured only to be initiated in historically validated market environments. This proactively avoids over 50% of ineffective trades in adverse environments such as "double-short counter-trend," thereby significantly improving the overall win rate and stability of the strategy from a macro perspective and avoiding unnecessary capital drawdowns.

[0088] Table 1

[0089] (2) By screening through technical factors, high-quality samples are selected in advance, which improves the purity and value density of subsequent AI model learning data.

[0090] Specifically, such as Figure 4 The image shown is a bar chart illustrating the average return of the Williams %R indicator over multiple periods over N days for stocks that have naturally hit their daily price limit, as provided in this application embodiment. Figure 5 The image shown is a bar chart illustrating the average return of a multi-period momentum indicator over N days for naturally limit-up stocks, as provided in this application embodiment. Figure 6 The image shown is a bar chart illustrating the average return on the turnover rate of stocks that have hit their daily price limit for N days under the natural limit-up price trend, as provided in this application embodiment. Figure 7The diagram shown is a bar chart illustrating the average return of the number of intraday limit-up days under N days of holding natural limit-up stocks provided in this application embodiment. The four technical factors used (multi-period Williams %R indicator, multi-period momentum indicator, limit-up day turnover rate, and intraday limit-up number) all exhibit clear monotonicity in their grouped returns, proving that these technical factors have a strong ability to identify high-quality limit-up stocks. For example, the holding returns of the lowest Williams %R group are significantly higher than those of other groups, verifying the effectiveness of the oversold rebound logic. It can be seen that the above four technical factors can respectively characterize the features of a high-quality limit-up stock with subsequent potential from different aspects such as oversold state, momentum exhaustion, capital consumption, and limit-up strength. Through a logical OR relationship, they form an efficient initial selection filter, excluding samples that are limit-up in form but lack subsequent potential in advance. This not only provides cleaner and higher-quality training data and prediction objects for downstream AI models, enabling AI models to focus more on learning high-value patterns, but also directly contributes to a higher average return of the final stock selection portfolio.

[0091] (3) By integrating multi-source features into AI prediction, a deep intelligent assessment of stock potential is achieved, ensuring the accuracy and directionality of the prediction signal.

[0092] Specifically, the input features of the AI ​​model integrate market state characteristics and technical factors along with their combined scores. This feature engineering ensures that the AI ​​model's decisions are based on a comprehensive assessment of the market environment and individual stock quality. This design transforms AI prediction from a black-box operation into deep reasoning built upon solid market logic and artificial intelligence. Figure 8 The diagram shows the net return curve of the Gradient Boosting Tree (GBT) model for predicting the next N trading days provided in this application embodiment. The portfolio constructed based on the model's predicted probability has a net return curve that significantly outperforms the market benchmark (CSI 300), especially in terms of 3-day, 5-day, and 10-day holding periods. This proves that AI prediction that integrates multi-source features can output stock selection signals with clear directionality and high practical value.

[0093] (4) The entire technical solution, through modular collaboration, ultimately forms a feasible investment strategy with a high success rate and high returns.

[0094] Specifically, through the connection and coordination of limit-up stock selection and AI prediction, market timing, initial sample selection, AI fine screening, and comprehensive ranking are organically integrated. The final comprehensive model achieved excellent performance in backtesting over nearly 5 years, with a 3-day return of 1.36% and a win rate of 53.67%; and a 20-day return of 5.69% and a win rate of 57.43%, as shown in Table 2 below. The win rate consistently exceeded 50% across all holding periods, with significant returns. It is evident that through the creative connection and data flow logic of each core step, the investment objective of initiating a stock selection strategy under suitable market trading conditions was achieved, ultimately translating into stable and outstanding investment performance.

[0095] Table 2

[0096] This application also proposes a data processing apparatus. Figure 9 This is a schematic diagram of the structure of the data processing device provided in the embodiments of this application, as shown below. Figure 9 As shown, the device includes: The filtering unit 901 is used to filter out a set of stocks that have hit the daily limit from historical data; to filter the set of stocks that have hit the daily limit based on market status, thereby obtaining a first candidate set of stocks; and to filter the first candidate set of stocks based on technical factors, thereby obtaining a second candidate set of stocks.

[0097] The prediction unit 902 is used to input the second candidate stock set into the network model for intelligent prediction and obtain the predicted profit probability list corresponding to the second candidate stock set output by the network model.

[0098] The screening unit 901 is also used to select a target stock set from the second candidate stock set based on the predicted profit probability list.

[0099] In some embodiments, historical data includes trading data for multiple stocks on multiple first trading days; wherein... Filtering unit 901 is specifically used for: For each of the multiple first trading days, obtain the closing price and opening price of each stock in the multiple stocks on each first trading day. If there are one or more stocks that have reached the daily limit price on each first trading day and whose closing price is equal to the daily limit price and whose opening price is lower than the closing price, then select one or more stocks that have reached the daily limit price from the multiple stocks on each first trading day, and determine the set of stocks that have reached the daily limit price based on one or more stocks that have reached the daily limit price.

[0100] In some embodiments, the filtering unit 901 is further specifically used for: Get the market index of each stock that hits its daily limit in the set of stocks that hit their daily limit. The market index of each stock that hit the daily limit was calculated in the first and second periods respectively, and the first smoothed moving average of the difference MACD indicator and the second MACD indicator corresponding to each stock that hit the daily limit were obtained. For each stock that hits its daily limit up, if both the first MACD indicator and the second MACD indicator corresponding to that stock are in the bullish zone, then that stock is designated as the first candidate stock; or... If the first MACD indicator corresponding to the stock that hits its daily limit is in the downtrend and the second MACD indicator corresponding to the stock that hits its daily limit is in the uptrend, then the stock that hits its daily limit is identified as the first candidate stock.

[0101] In some embodiments, the filtering unit 901 is further specifically used for: Obtain the preset set of technical factors; The first candidate stock set is screened using technical factors to obtain the second candidate stock set.

[0102] In some embodiments, the set of technical factors includes one or more of the following: multi-period Williams %R percentile, multi-period trend amplitude, daily turnover rate on limit-up days, and cumulative number of times the stock hits the limit-up price during trading; wherein... The filtering unit 901 is also specifically used for one of the following: For each first candidate stock in the first candidate stock set, indicators are calculated for each first candidate stock over multiple periods to obtain multiple Williams indicators for each first candidate stock; the multiple Williams indicators for each first candidate stock are summed to obtain the total Williams indicator for each first candidate stock; if there are one or more first candidate stocks with a total Williams indicator less than a first threshold, then one or more first candidate stocks are determined as one or more second candidate stocks. For each first candidate stock, the indicators are calculated over multiple periods to obtain multiple momentum indicators corresponding to each first candidate stock; the multiple momentum indicators corresponding to each first candidate stock are summed to obtain the total momentum indicator corresponding to each first candidate stock; if there are one or more first candidate stocks with a total momentum indicator less than a second threshold, then one or more first candidate stocks are determined as one or more second candidate stocks. Obtain the turnover rate of each first candidate stock on the first trading day; if there are one or more first candidate stocks with a turnover rate less than the third threshold, then determine one or more first candidate stocks as one or more second candidate stocks. Obtain the number of limit-up days for each first candidate stock on the first trading day. If there are one or more first candidate stocks with a number of limit-up days greater than the fourth threshold, then determine one or more first candidate stocks as one or more second candidate stocks.

[0103] In some embodiments, the prediction unit 902 is specifically used for: Based on the first MACD indicator and the second MACD indicator corresponding to each second candidate stock in the second candidate stock set, the market technical factor corresponding to each second candidate stock is determined. The overall score for each second candidate stock is determined based on the total Williams %R, total momentum %R, and the number of limit-up days on the first trading day. Based on one or more of the following factors for each second candidate stock: first MACD indicator, second MACD indicator, total Williams %R indicator, total momentum indicator, turnover rate on the first trading day, number of limit-up days on the first trading day, market technical factors, and comprehensive score, determine the set of characteristic factors corresponding to each second candidate stock. The feature factor set corresponding to each second candidate stock is input into the network model for intelligent prediction, and the predicted profit probability list output by the network model is obtained.

[0104] In some embodiments, the predicted profitability probability list includes the price change label and profitability probability of each second candidate stock over multiple second trading days; wherein... Filtering unit 901 is also specifically used for: Based on the price change label and profit probability of each second candidate stock in multiple second trading days, the set of second candidate stocks is sorted in descending order to obtain the sorted set of second candidate stocks. Select a preset number of second candidate stocks from the sorted second candidate stock set, and determine the target stock set based on the preset number of second candidate stocks.

[0105] Those skilled in the art should understand that Figure 9 The functions of each unit in the data processing device shown can be understood by referring to the relevant description of the aforementioned method. Figure 9 The functions of each unit in the data processing device shown can be implemented by a program running on a processor or by specific logic circuits.

[0106] Figure 10 This is a schematic diagram of the processing device provided in an embodiment of this application. The processing device may be a terminal device or a network device. Figure 10 The processing device shown includes a processor 1001, which can call and run computer programs from memory to implement the methods in the embodiments of this application.

[0107] Optionally, such as Figure 10 As shown, the processing device may further include a memory 1002. The processor 1001 can retrieve and run computer programs from the memory 1002 to implement the methods described in the embodiments of this application.

[0108] The memory 1002 can be a separate device independent of the processor 1001, or it can be integrated into the processor 1001.

[0109] Optionally, such as Figure 10 As shown, the processing device may also include a transceiver 1003, which the processor 1001 can control to communicate with other devices. Specifically, it can send information or data to other devices or receive information or data sent by other devices.

[0110] The transceiver 1003 may include a transmitter and a receiver. The transceiver 1003 may further include an antenna, and the number of antennas may be one or more.

[0111] The processing device may specifically be the processing device of the embodiments of this application, and the processing device can implement the corresponding processes of the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.

[0112] It should be understood that the processor in the embodiments of this application may be an integrated circuit chip with signal processing capabilities. In implementation, the steps of the above method embodiments can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor described above can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.

[0113] It is understood that the memory in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus RAM (DR RAM). It should be noted that the memory used in the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0114] This application also provides a computer-readable storage medium for storing a computer program. This computer-readable storage medium can be applied to the processing device in this application embodiment, and the computer program causes the computer to execute the corresponding processes implemented by the various methods in this application embodiment; for brevity, further details are omitted here.

[0115] This application also provides a computer program product, including computer program instructions. This computer program product can be applied to the processing device in this application embodiment, and the computer program instructions cause the computer to execute the corresponding processes implemented by the various methods in this application embodiment; for brevity, further details are omitted here.

[0116] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0117] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0118] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0119] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0120] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0121] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0122] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A data processing method, characterized by, The method includes: Select a set of stocks that hit the daily limit up from historical data; The market status of the set of stocks that hit the daily limit is filtered to obtain the first candidate stock set; The first candidate stock set is screened using technical factors to obtain the second candidate stock set; The second candidate stock set is input into the network model for intelligent prediction, and the predicted profit probability list corresponding to the second candidate stock set output by the network model is obtained. The target stock set is selected from the second candidate stock set based on the predicted profit probability list.

2. The method of claim 1, wherein, The historical data includes trading data for multiple stocks on multiple first trading days; The set of stocks that hit their daily limit up, selected from historical data, includes: For each of the plurality of first trading days, obtain the closing price and opening price of each of the plurality of stocks on each first trading day. If there are one or more stocks that have reached the daily limit price on each first trading day, and the closing price is equal to the daily limit price and the opening price is lower than the closing price, then select the one or more stocks that have reached the daily limit price from the plurality of stocks on each first trading day, and determine the set of stocks that have reached the daily limit price based on the one or more stocks that have reached the daily limit price.

3. The method according to claim 1, characterized in that, The process of filtering the set of stocks that have hit their daily limit up based on market conditions to obtain a first candidate stock set includes: Obtain the market index of each stock that hits its daily limit in the set of stocks that hit their daily limit; The market index of each stock that hit the daily limit is calculated in the first period and the second period respectively to obtain the first smoothed moving average MACD indicator and the second MACD indicator corresponding to each stock that hit the daily limit. For each stock that hits its daily limit, if both the first MACD indicator and the second MACD indicator corresponding to that stock are in the bullish zone, then that stock is designated as the first candidate stock; or... If the first MACD indicator corresponding to the stock that hits its daily limit is in the downtrend and the second MACD indicator corresponding to the stock that hits its daily limit is in the uptrend, then the stock that hits its daily limit is identified as the first candidate stock.

4. The method according to claim 1, characterized in that, The step of performing technical factor screening on the first candidate stock set to obtain the second candidate stock set includes: Obtain the preset set of technical factors; Based on the set of technical factors, the first candidate stock set is screened using technical factors to obtain the second candidate stock set.

5. The method according to claim 4, characterized in that, The set of technical factors includes one or more of the following: multi-period Williams %R, multi-period trend amplitude, turnover rate on limit-up days, and cumulative number of times the limit-up price is touched during the trading session. The second candidate stock set is obtained by screening the first candidate stock set based on the set of technical factors, and includes one of the following: For each first candidate stock in the first candidate stock set, indicators are calculated for each first candidate stock over multiple periods to obtain multiple Williams indicators corresponding to each first candidate stock. The Williams %R values ​​corresponding to each first candidate stock are summed to obtain the total Williams %R value corresponding to each first candidate stock. If there are one or more first candidate stocks whose total Williams %R is less than the first threshold, then the one or more first candidate stocks will be determined as one or more second candidate stocks. The indicators for each first candidate stock are calculated over multiple periods to obtain multiple momentum indicators corresponding to each first candidate stock. The total momentum index corresponding to each first candidate stock is obtained by summing up the multiple momentum indices corresponding to each first candidate stock. If there are one or more first candidate stocks with a total momentum index less than the second threshold, then the one or more first candidate stocks are determined as one or more second candidate stocks; Obtain the turnover rate of each first candidate stock on the first trading day; If there are one or more first candidate stocks with a turnover rate less than the third threshold, then the one or more first candidate stocks will be determined as one or more second candidate stocks; Obtain the number of limit-up days for each first candidate stock on the first trading day. If there are one or more first candidate stocks with a number of limit-up days greater than the fourth threshold, then the one or more first candidate stocks are determined as one or more second candidate stocks.

6. The method according to claim 5, characterized in that, The step of inputting the second candidate stock set into the network model for intelligent prediction, and obtaining the predicted profit probability list corresponding to the second candidate stock set output by the network model, includes: Based on the first MACD indicator and the second MACD indicator corresponding to each second candidate stock in the second candidate stock set, the market technical factor corresponding to each second candidate stock is determined. Based on the total Williams %R, total momentum %R, and number of limit-up days on the first trading day for each second candidate stock, a comprehensive score is determined for each second candidate stock. Based on one or more of the following indicators for each second candidate stock: first MACD indicator, second MACD indicator, total Williams %R indicator, total momentum indicator, turnover rate on the first trading day, number of limit-up days on the first trading day, market technical factors, and comprehensive score, determine the set of characteristic factors corresponding to each second candidate stock. The feature factor set corresponding to each second candidate stock is input into the network model for intelligent prediction, and the predicted profit probability list output by the network model is obtained.

7. The method according to any one of claims 1 to 6, characterized in that, The predicted profitability probability list includes the price change label and profitability probability of each second candidate stock over multiple second trading days; The step of selecting the target stock set from the second candidate stock set based on the predicted profitability probability list includes: Based on the price change label and profit probability of each second candidate stock in multiple second trading days, the second candidate stock set is sorted in descending order to obtain the sorted second candidate stock set. Select a preset number of second candidate stocks from the sorted second candidate stock set, and determine the target stock set based on the preset number of second candidate stocks.

8. A data processing apparatus, characterized in that, The device includes: A filtering unit is used to filter out a set of stocks that have hit their daily limit from historical data; to filter the set of stocks that have hit their daily limit based on market status to obtain a first candidate stock set; and to filter the first candidate stock set based on technical factors to obtain a second candidate stock set. The prediction unit is used to input the second candidate stock set into the network model for intelligent prediction and obtain the predicted profit probability list corresponding to the second candidate stock set output by the network model. The filtering unit is further configured to filter out a target stock set from the second candidate stock set based on the predicted profit probability list.

9. A processing device, characterized in that, include: A processor and a memory for storing a computer program, the processor for calling and running the computer program stored in the memory to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Used to store a computer program that causes a computer to perform the method as described in any one of claims 1 to 7.

11. A computer program product, characterized in that, It includes computer program instructions that cause a computer to perform the method as described in any one of claims 1 to 7.

Images