Stock quantification method based on multi-dimensional quantitative indexes

Abstract

The invention relates to a stock quantification method based on a multi-dimensional quantification index, and belongs to the technical field of data analysis, and the stock quantification method based on the multi-dimensional quantification index comprises the following steps: rapid stock selection, parameter optimization, signal generation and quarterly updating. Through a dynamic parameter optimization mechanism, real-time response to market short-term inertia and medium-and-long-term trend is realized, flexibility and robustness of a transaction strategy can be maintained in different market environments, a high-quality stock pool is rapidly screened by using multi-dimensional quantitative indexes, the yield and risk are balanced, and the market quality is improved. According to the method, minute-level screening of stocks in the whole market is realized, a genetic algorithm optimization technology is introduced, an optimal solution is quickly searched in various parameter combinations, the yield is remarkably improved, a dynamic optimization strategy is combined, it is ensured that parameters are adjusted in real time along with market inertia changes, local optimum is avoided, and accurate transaction decision support is realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of data analysis, and particularly relates to a stock quantification method based on multi-dimensional quantification indicators. Background Art

[0002] Quantitative trading strategies are important tools in the financial field. Their core lies in using data and algorithms to formulate trading plans to reduce the influence of human emotions and improve efficiency. In today's rapidly changing stock market, market participants have higher and higher requirements for trading strategies, demanding better dynamic adjustment capabilities and market adaptability. Quantitative trading strategies have experienced rapid development in recent years, especially driven by data processing capabilities and artificial intelligence technologies. The latest development trends in this field are as follows: High-frequency trading: High-frequency trading is an important branch of quantitative trading, which makes rapid decisions by leveraging market fluctuations at the microsecond level, posing extremely high requirements for the algorithm efficiency, execution speed, and trading cost control of trading strategies. However, it generally relies on fixed parameters and has limited performance in the face of unstructured market fluctuations. Deep learning models are gradually applied in quantitative trading for price prediction, risk management, and portfolio optimization. For example, LSTM and Transformer models can capture complex patterns in time series data. Reinforcement learning provides new solutions for intelligent trading decisions, especially performing well in multi-asset optimization strategies in dynamic market environments. The diversity and real-time nature of data bring new opportunities to quantitative trading, such as using non-traditional data sources like social media sentiment and news hotspots. The expansion of cloud computing enables the processing of massive data and the training of complex models, further lowering the technical threshold. The introduction of emerging technologies such as blockchain in the financial field has led to the rise of decentralized trading strategies, further expanding the application scenarios of quantitative trading.

[0003] Traditional quantitative trading strategies are usually based on simple technical indicators or fixed rules. However, in a complex and rapidly changing market, these methods cannot adapt to the dynamic changes of the market in a timely manner, which may lead to increased volatility of returns and missed market opportunities. Moreover, the rapid development of the above technologies has brought unprecedented opportunities, but also put forward higher requirements for the dynamic adaptability and robustness of strategies. Traditional fixed-parameter models gradually expose their limitations when facing a complex and rapidly changing market. Based on this, a stock quantification method based on multi-dimensional quantification indicators is proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide a reasonably designed stock quantification method based on multi-dimensional quantification indicators to solve the above problems.

[0005] The present invention achieves the above purpose through the following technical solutions:

[0006] A stock quantitative method based on multi-dimensional quantitative indicators includes the following steps:

[0007] Step 1: Obtain ChiNext stock data, screen the stock data through multi-dimensional quantitative indicators, assign different weight coefficients to the multi-dimensional quantitative indicators, calculate the comprehensive score, and select high-score stocks to construct a stock pool;

[0008] Step 2: Construct a parameter optimization model based on the genetic algorithm, and optimize the short-term moving average parameter, long-term moving average parameter, short-term VWAP parameter, long-term VWAP parameter, and position opening and closing coefficients through the parameter optimization model to obtain a hyperparameter combination with a high backtest return rate;

[0009] Step 3: Based on the hyperparameter combination, use the parameter optimization model to generate trading signals through the daily market data of the stocks in the stock pool as a reference for the next day's trading;

[0010] Step 4: Update the stock pool and the hyperparameter combination at the end of each quarter to dynamically quantify the stock data.

[0011] As a further optimization scheme of the present invention, the multi-dimensional quantitative indicators include: Alpha, Beta, Sharpe ratio, and logarithmic cumulative return;

[0012] The mathematical expression of Alpha is: α = (Ri - Rf) - β(Rm - Rf)

[0013] Where Ri is the actual return of the asset;

[0014] Rf is the risk-free return;

[0015] Rm is the market benchmark return;

[0016] β is the market risk coefficient of the asset;

[0017] The mathematical expression of Beta:

[0018] Where Cov(Ri, Rm) is the covariance between the actual return of the asset and the market benchmark return;

[0019] Var(Rm) is the variance of the market benchmark return;

[0020] is the average value of the actual return of the asset;

[0021] is the average value of the market benchmark return;

[0022] t is time;

[0023] The mathematical expression of the Sharp ratio:

[0024] Among them, Ri is the actual return rate of the asset;

[0025] Rf is the risk-free return rate;

[0026] σi is the standard deviation of the actual return rate of the asset;

[0027] Mathematical expression of logarithmic cumulative return rate:

[0028] Among them, Pt is the asset price at time t;

[0029] Pt-1 is the asset price at time t-1.

[0030] As a further optimization scheme of the present invention, the comprehensive score calculation formula is as follows:

[0031] Comprehensive score = weight coefficient 1 * Alpha + weight coefficient 2 * Sharpe ratio + weight coefficient 3 * logarithmic cumulative return rate + weight coefficient 4 * (1 - ABS(Beta - 1));

[0032] The specific steps for calculating the comprehensive score are as follows:

[0033] Read the historical data of all stocks;

[0034] Calculate the Alpha, Beta, Sharpe ratio and logarithmic cumulative return rate of each stock;

[0035] Calculate the comprehensive score, and based on the comprehensive score data, screen out multiple stocks and output them as the stock pool for the current quarter.

[0036] As a further optimization scheme of the present invention, a parameter optimization model is constructed based on the genetic algorithm, and the short-term moving average parameter, long-term moving average parameter, short-term VWAP parameter, long-term VWAP parameter, and position opening and closing coefficients are optimized through the parameter optimization model to obtain a hyperparameter combination with a high backtest return rate. The specific optimization steps are as follows:

[0037] Read the historical market data of each stock in the stock pool;

[0038] Calculate and generate multiple parameter combinations through the parameter optimization model constructed by the genetic algorithm;

[0039] Optimize multiple parameter combinations to obtain a hyperparameter combination with a high backtest return rate;

[0040] Output the optimization result and save it to the parameter summary file.

[0041] As a further optimization scheme of the present invention, the steps to obtain a hyperparameter combination with a high backtest return rate are as follows:

[0042] Set parameters;

[0043] Generate trading signals based on the parameters;

[0044] Close positions according to the trading signals;

[0045] Determine the rate of return and obtain the hyperparameter combination with a high backtest rate of return.

[0046] As a further optimization scheme of the present invention, the parameter setting includes moving average lines for identifying trends, volume-weighted average prices that combine short-term VWAP and long-term VWAP and are used to judge the direction of market fund flow, test directions, and parameter settings of entry_coef and exit_coef for dynamically adjusting the entry and exit price trigger conditions;

[0047] Among them, the moving average lines include the 3-day moving average line, the 5-day moving average line, short-term moving average lines, and long-term moving average lines.

[0048] As a further optimization scheme of the present invention, trading signals are generated based on the parameters, and the trading signals include long signals and short signals;

[0049] The generation conditions for long signals are: when the 5-day moving average line ≥ short-term moving average lines, the 5-day moving average line ≥ the specified amplitude of the short-term moving average lines, short-term VWAP > long-term VWAP, and the current position is 0, a long signal is generated;

[0050] Among them, the calculation method for the specified amplitude of the short-term moving average lines: short-term moving average lines + (ABS(short-term moving average lines - long-term moving average lines) * entry_coef);

[0051] The generation conditions for short signals are: when the 5-day moving average line ≤ short-term moving average lines, the 5-day moving average line ≤ the specified amplitude of the short-term moving average lines, short-term VWAP < long-term VWAP, and the current position is 0, a short signal is generated;

[0052] Among them, the calculation method for the specified amplitude of the short-term moving average lines: short-term moving average lines - (ABS(short-term moving average lines - long-term moving average lines) * entry_coef).

[0053] As a further optimization scheme of the present invention, closing positions according to the trading signals includes long position closing and short position closing;

[0054] Among them, long - position closing includes step - by - step closing and full - position closing. When the 3 - day moving average ≤ 5 - day moving average and the short - term VWAP < long - term VWAP, the first step - by - step closing is carried out, and 50% of the position is closed. When the 3 - day moving average ≤ 5 - day moving average, the 3 - day moving average ≤ the specified range of the 5 - day moving average, and the short - term VWAP < long - term VWAP, the second step - by - step closing is carried out at this time, and 50% of the position is closed. Among them, the specified range of the 5 - day moving average = 5 - day moving average-(ABS(5 - day moving average - short - term moving average)*exit_coef);

[0055] Full - position closing includes: when the stop - loss condition, that is, the holding loss range ≥ 5%, full - position closing is carried out;

[0056] When the target condition, that is, the short - term VWAP < long - term VWAP, full - position closing is carried out;

[0057] Short - position closing is divided into step - by - step closing and full - position closing. When the 3 - day moving average ≥ 5 - day moving average and the short - term VWAP > long - term VWAP, the first step - by - step closing is carried out, and 50% of the position is closed. When the 3 - day moving average ≥ 5 - day moving average, the 3 - day moving average ≥ the specified range of the 5 - day moving average, and the short - term VWAP > long - term VWAP, the second step - by - step closing is carried out at this time, and 50% of the position is closed. Among them, the specified range of the 5 - day moving average = 5 - day moving average+(ABS(5 - day moving average - short - term moving average)*exit_coef);

[0058] Full - position closing includes: when the stop - loss condition, that is, the holding loss range ≥ 5%, full - position closing is carried out;

[0059] When the target condition, that is, the short - term VWAP > long - term VWAP, full - position closing is carried out.

[0060] As a further optimization scheme of the present invention, based on the hyper - parameter combination, and using the parameter optimization model to generate trading signals through the daily market data of the stocks in the stock pool as the trading reference for the next day. The specific steps are as follows:

[0061] Update the market data of each stock in the stock pool;

[0062] Generate trading signals through the parameter optimization model;

[0063] Output and save the information file according to the trading information.

[0064] As a further optimization scheme of the present invention, at the end of each quarter, update the stock pool and the hyper - parameter combination to dynamically quantify stock data. The specific steps are as follows:

[0065] Repeat Step 1, Step 2, and Step 3 to update the stock pool and the hyper - parameter combination of the previous quarter;

[0066] Monitor the open positions until the positions are closed;

[0067] Output stock quantitative data.

[0068] The beneficial effects of the present invention are as follows:

[0069] 1. Through the dynamic parameter optimization mechanism, the present invention realizes real-time response to the short-term inertia and medium- and long-term trends of the market, and can maintain the flexibility and robustness of trading strategies in different market environments. By using multi-dimensional quantitative indicators, it quickly screens high-quality stocks to construct a high-quality stock pool, balances the return rate and risk, realizes minute-level screening of all-market stocks, introduces genetic algorithm optimization technology, quickly searches for the optimal solution among various parameter combinations, significantly improves the return rate, and combines with the dynamic optimization strategy to ensure that the parameters are adjusted in real time according to the market inertia, avoiding falling into local optima. In addition, the market data is updated daily, and long-short trading signals are generated according to the optimized parameters, providing a reliable basis for the next-day trading. The trading rules based on VWAP are introduced to achieve precise trading decision support.

[0070] 2. By designing a dynamic stop-profit and stop-loss mechanism, the present invention precisely controls trading risks through the changes in moving averages and parameter combinations. During the quarterly update process, the unclosed stocks are smoothly connected to ensure the continuity of risk management. The functions such as rapid stock selection, parameter optimization, signal generation, and quarterly update are integrated into an integrated framework, reducing human intervention and realizing automated execution of the strategy. And through quarterly dynamic update, it ensures that the strategy always follows the market rhythm. Description of the Drawings

[0071] Figure 1 is the overall process schematic diagram of the present invention;

[0072] Figure 2 is the comparison chart of the logarithmic return rate of sz300502 stock of the present invention, the return rate of the ChiNext Index, and the return rate of treasury bonds;

[0073] Figure 3 is the comparison chart of the logarithmic cumulative return rate of sz300502 stock of the present invention, the cumulative return rate of the ChiNext Index, and the cumulative return rate of treasury bonds;

[0074] Figure 4 is the comparison chart of the logarithmic return rate of sz300563 stock of the present invention, the return rate of the ChiNext Index, and the return rate of treasury bonds;

[0075] Figure 5 is the comparison chart of the logarithmic cumulative return rate of sz300563 stock of the present invention, the cumulative return rate of the ChiNext Index, and the cumulative return rate of treasury bonds. Detailed Embodiments

[0076] The present application will be further described in detail below with reference to the accompanying drawings. It is necessary to point out here that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the protection scope of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0077] Example 1: As Figure 1 shown, a stock quantification method based on multi-dimensional quantification indexes includes the following steps:

[0078] Step 1: Quick stock selection: Obtain the data of ChiNext stocks, screen the stock data through multi-dimensional quantification indexes, assign different weight coefficients to the multi-dimensional quantification indexes, calculate the comprehensive score, and select high-score stocks to construct a stock pool. The multi-dimensional quantification indexes include: Alpha, Beta, Sharpe ratio, and logarithmic cumulative return rate;

[0079] The mathematical expression of Alpha is: α = (Ri - Rf) - β(Rm - Rf)

[0080] where Ri is the actual return rate of the asset;

[0081] Rf is the risk-free return rate;

[0082] Rm is the market benchmark return rate;

[0083] β is the market risk coefficient of the asset;

[0084] The mathematical expression of Beta:

[0085] where Cov(Ri, Rm) is the covariance between the actual return rate of the asset and the market benchmark return rate;

[0086] Var(Rm) is the variance of the market benchmark return rate;

[0087] is the average value of the actual return rate of the asset;

[0088] is the average value of the market benchmark return rate;

[0089] t is time;

[0090] The mathematical expression of the Sharp ratio:

[0091] where Ri is the actual return rate of the asset;

[0092] Rf is the risk-free return rate (one-year treasury bond yield);

[0093] σi is the standard deviation of the actual return rate of the asset;

[0094] Mathematical expression of logarithmic cumulative return rate:

[0095] Wherein, Pt is the asset price at time t;

[0096] Pt-1 is the asset price at time t-1, and the comprehensive score is calculated by assigning different weight coefficients to multi-dimensional quantitative indicators;

[0097] The calculation formula of the comprehensive score is as follows:

[0098] Comprehensive score = weight coefficient 1 * Alpha + weight coefficient 2 * Sharpe ratio + weight coefficient 3 * logarithmic cumulative return rate + weight coefficient 4 * (1 - ABS(Beta - 1));

[0099] The specific steps for calculating the comprehensive score are as follows:

[0100] S101: Read the historical data of all stocks;

[0101] S102: Calculate the Alpha, Beta, Sharpe ratio and logarithmic cumulative return rate of each stock;

[0102] S103: Calculate the comprehensive score, and screen out multiple stocks according to the comprehensive score data, and output them as the stock pool for the current quarter;

[0103] In the implementation of Step 1, multi-dimensional quantitative indicators are used to ensure that the stock selection results take into account the return rate, stability and market volatility risk, and the efficiency is significantly improved;

[0104] Step 2: Parameter optimization: Based on the genetic algorithm, a parameter optimization model is constructed, and the short-term moving average parameter, long-term moving average parameter, short-term VWAP parameter, long-term VWAP parameter (VWAP is the volume-weighted average price), and the position opening and closing coefficients are optimized through the parameter optimization model to obtain a hyperparameter combination with a high backtest return rate. The specific optimization steps are as follows:

[0105] S201: Read the historical market data of each stock in the stock pool;

[0106] S202: Calculate and generate multiple parameter combinations through the parameter optimization model constructed by the genetic algorithm;

[0107] S203: The steps for optimizing multiple parameter combinations to obtain a hyperparameter combination with a high backtest return rate are as follows:

[0108] S20301: Set parameters including moving averages for identifying trends, volume-weighted average prices (VWAP) that combine short-term and long-term VWAP and are used to judge the direction of market fund flow, the test direction (selectable as long, short, or both), and the parameter settings of entry_coef (position establishment coefficient) and exit_coef (position closing coefficient) for dynamically adjusting the entry and exit price trigger conditions. Among them, the moving averages include the 3-day moving average, 5-day moving average, short-term moving average, and long-term moving average. VWAP is the volume-weighted average price. The parameter search range is as follows:

[0109] Short-term moving average range: short_ma = random.randint(9, 21);

[0110] Long-term moving average range: long_ma = random.randint(29, 91);

[0111] Short-term VWAP range: short_vwap = random.randint(5, 10);

[0112] Long-term VWAP range: long_vwap = random.randint(11, 45);

[0113] entry_coef range: entry_coef = round(random.uniform(0.24, 0.50), 2);

[0114] exit_coef range: exit_coef = round(random.uniform(0.17, 0.33), 2);

[0115] The above parameter combination possibilities are 78,943,410 kinds. Calculate the optimal solution within 78,943,410 kinds through the genetic algorithm;

[0116] S20302: Generate trading signals according to the parameters. The trading signals include long signals and short signals;

[0117] The generation condition for the long signal is: when the 5-day moving average ≥ short-term moving average, the 5-day moving average ≥ the specified amplitude of the short-term moving average, short-term VWAP > long-term VWAP, and the current position is 0, generate a long signal;

[0118] Among them, the calculation method for the specified amplitude of the short-term moving average: short-term moving average + (ABS(short-term moving average - long-term moving average) * entry_coef);

[0119] The operation is: buy 100 units at the opening price on the next trading day;

[0120] The short signal generation conditions are as follows: When the 5-day moving average ≤ the short-term moving average, the 5-day moving average ≤ the specified range of the short-term moving average, the short-term VWAP < the long-term VWAP, and the current position is 0, a short signal is generated;

[0121] Among them, the calculation method of the specified range of the short-term moving average is: short-term moving average - (ABS(short-term moving average - long-term moving average) * entry_coef), where ABS is the absolute value;

[0122] The operation is: sell 100 units at the opening price on the next trading day;

[0123] S20303: Close positions according to the trading signal, and position closing includes long position closing and short position closing;

[0124] Among them, long position closing includes step-by-step position closing and full position closing. When the 3-day moving average ≤ the 5-day moving average and the short-term VWAP < the long-term VWAP, the first step of step-by-step position closing is carried out, and 50% of the position is closed. When the 3-day moving average ≤ the 5-day moving average, the 3-day moving average ≤ the specified range of the 5-day moving average, and the short-term VWAP < the long-term VWAP, the second step of step-by-step position closing is carried out at this time, and 50% of the position is closed. Among them, the specified range of the 5-day moving average = 5-day moving average - (ABS(5-day moving average - short-term moving average) * exit_coef);

[0125] Full position closing includes: When the stop-loss condition, that is, the holding loss range ≥ 5%, full position closing is carried out;

[0126] When the target condition, that is, the short-term VWAP < the long-term VWAP, full position closing is carried out;

[0127] Short position closing is divided into step-by-step position closing and full position closing. When the 3-day moving average ≥ the 5-day moving average and the short-term VWAP > the long-term VWAP, the first step of step-by-step position closing is carried out, and 50% of the position is closed. When the 3-day moving average ≥ the 5-day moving average, the 3-day moving average ≥ the specified range of the 5-day moving average, and the short-term VWAP > the long-term VWAP, the second step of step-by-step position closing is carried out at this time, and 50% of the position is closed. Among them, the specified range of the 5-day moving average = 5-day moving average + (ABS(5-day moving average - short-term moving average) * exit_coef);

[0128] Full position closing includes: When the stop-loss condition, that is, the holding loss range ≥ 5%, full position closing is carried out;

[0129] When the target condition, that is, the short-term VWAP > the long-term VWAP, full position closing is carried out;

[0130] S20304: Determine the rate of return and obtain the hyperparameter combination with a high backtest rate of return;

[0131] It should be noted that the specific steps shown in S203 are the steps for constructing the parameter optimization model, and the specific steps within S203 are run within the parameter optimization model. Moreover, the optimization parameter runs based on the fixed values of the hyperparameters already found by the parameter optimization model to generate trading signals;

[0132] S204: Output the optimization result and save it to the parameter summary file;

[0133] The implementation of Step 2 overcomes the limitation of parameter selection relying on experience in traditional methods through global search, and dynamically optimizes parameters to ensure that the strategy adapts to market inertia changes;

[0134] Step 3: Signal generation: Based on the hyperparameter combination, and using the parameter optimization model to generate trading signals through the daily market data of the stocks in the stock pool as a reference for the next day's trading. The specific steps are as follows:

[0135] S301: Update the market data of each stock in the stock pool;

[0136] S302: Generate trading signals through the parameter optimization model;

[0137] S303: Output and save the information file according to the trading information;

[0138] The implementation of Step 3 realizes the rolling update of data, ensures the timeliness and accuracy of trading signals, and introduces trading rules based on VWAP to achieve precise trading decision support;

[0139] Step 4: Quarterly update: Update the stock pool and hyperparameter combination at the end of each quarter, and dynamically quantify stock data. The specific steps are as follows:

[0140] S401: Repeat Step 1, Step 2, and Step 3 to update the stock pool and hyperparameter combination of the previous quarter;

[0141] S402: Monitor the open positions of stocks until they are closed;

[0142] S403: Output the stock quantification data;

[0143] The implementation of Step 4 combines market dynamic adjustment strategies, improves long-term return performance, and realizes seamless connection between the strategy and market inertia.

[0144] Example 2: As Figure 1 、 Figure 2 、 Figure 3 、 Figure 4 and Figure 5As shown in the figure, Step 1: Run the stock selection algorithm based on the historical data of 1,363 stocks on the ChiNext from January 1, 2022 to December 20, 2024. The output result file includes the stock name, Alpha value, Beta value, Sharpe ratio, and cumulative return rate;

[0145] Its comprehensive score = 0.4 * Alpha + 0.2 * Sharpe ratio + 0.3 * logarithmic cumulative return rate + 0.1 * (1 - ABS(Beta - 1)). The comprehensive score results are shown in Table 1 (in descending order);

[0146] Table 1 Comprehensive scores of each stock

[0147]

[0148]

[0149] And select the top 10 stocks with the highest comprehensive scores according to Table 1 and put them into the stock pool;

[0150] Among them, the comparison charts of the stock return rates of sz300502 and sz300563 with the ChiNext index and the national debt return rate are as shown in Figure 2 , Figure 3 , Figure 4 and Figure 5 shown;

[0151] Step 2: Use the genetic algorithm to calculate the parameter combinations of the moving average, VWAP, and trading coefficient for the stocks selected into the stock pool, and take the parameters corresponding to the highest return rate as the hyperparameter combination;

[0152] For example, calculate the hyperparameters of stocks 300502 and 300563, and the calculation results are shown in Table 2 and Table 3;

[0153] Table 2 Calculation results of hyperparameters of stock 300502

[0154]

[0155]

[0156] Table 3 Calculation results of hyperparameters of stock 300563

[0157]

[0158]

[0159]

[0160] It should be noted that the profit calculation formulas in Table 2 and Table 3 are as follows: Profit = (selling price - buying price) * quantity;

[0161] And according to the trading strategy, there are two cases: (1) One buy corresponds to two sells; (2) When the loss of one buy is greater than 5%, sell all at once.

[0162] For example: For stock 300502, as can be seen from the data in Table 2, trading signals are generated based on the moving average and position - building coefficient calculated according to the closing price on June 8, 2022. Buy at the opening price on June 9, 2022, with the opening price of 25.08 and the number of shares bought being 100; According to the closing price on July 4, 2022, the calculated loss is greater than 5%, and sell at the opening price on July 5, 2022, with the opening price of 24.13 and the number of shares sold being 100.

[0163] Therefore, the profit obtained when selling on July 5, 2022 = (24.13 - 25.08) * 100 = - 167, and the negative profit actually means a loss.

[0164] For stock 300502, as can be seen from the data in Table 2, trading signals are generated based on the moving average and position - building coefficient calculated according to the closing price on January 27, 2023. Buy at the opening price on January 30, 2023, with the opening price of 25.51 and the number of shares bought being 100. According to the moving average calculated according to the closing price on June 12, 2023, a trading signal is generated, and sell at the opening price on June 13, 2023, with the opening price of 61.05 and the number of shares sold being 50. According to the moving average and closing - position coefficient calculated according to the closing price on June 27, 2023, a trading signal is generated, and sell at the opening price on June 28, 2023, with the opening price of 64.28 and the number of shares sold being 50.

[0165] Therefore, the profit obtained when selling on June 28, 2023 = ((61.50 - 25.51) * 50)+((64.28 - 25.51) * 50)=3738, and the positive profit actually means a gain.

[0166] For stock 300563, as can be seen from the data in Table 3, trading signals are generated based on the moving average and position - building coefficient calculated according to the closing price on February 22, 2022. Buy at the opening price on February 23, 2022, with the opening price of 14.53 and the number of shares bought being 100; According to the closing price on March 7, 2022, the calculated loss is greater than 5%, and sell at the opening price on March 8, 2022, with the opening price of 13.73 and the number of shares sold being 100.

[0167] Therefore, the profit obtained when selling on March 8, 2022 = (13.73 - 14.53) * 100 = - 80, and the negative profit actually means a loss.

[0168] For stock 300563, as can be seen from the data in Table 3, trading signals are generated based on the moving averages and position - building coefficients calculated according to the closing price on January 5, 2023. On January 6, 2023, it is bought at the opening price of 11.32, and the quantity bought is 100. Trading signals are generated based on the moving averages calculated according to the closing price on March 19, 2023. On March 20, 2023, it is sold at the opening price of 14, and the quantity sold is 50. Trading signals are generated based on the moving averages and position - closing coefficients calculated according to the closing price on March 20, 2023. On March 21, 2023, it is sold at the opening price of 14.05, and the quantity sold is 50;

[0169] Therefore, the profit obtained when selling on March 21, 2023 = ((14 - 11.32) * 50)+((14.05 - 11.32) * 50)=270.5. A positive profit means actual profit;

[0170] From the above results, it can be seen that the hyper - parameter combinations for each stock are different. After recording the hyper - parameters of each stock, they are used for daily update calculations in the next 3 months, forming a list of hyper - parameter combinations as shown in Table 4;

[0171] Table 4 Updated Hyper - parameter Combinations

[0172] stock 3DMA 5DMA short_ma long_ma short_vwap long_vwap entry_coef exit_coef sz300085 3 5 17 82 8 11 0.24 0.30 sz300100 3 5 18 44 5 25 0.44 0.33 sz300210 3 5 14 42 5 13 0.47 0.17 sz300290 3 5 25 56 10 12 0.26 0.26 sz300394 3 5 17 87 9 14 0.31 0.22 sz300489 3 5 7 46 8 16 0.41 0.19 sz300493 3 5 21 88 6 22 0.37 0.20 sz300502 3 5 21 65 9 44 0.42 0.20 sz300552 3 5 20 32 6 16 0.36 0.26 sz300561 3 5 27 91 5 17 0.32 0.17 sz300563 3 5 12 42 6 13 0.24 0.20 sz300570 3 5 8 57 10 25 0.40 0.17 sz300641 3 5 18 34 10 12 0.24 0.17 sz300757 3 5 8 69 9 11 0.47 0.18

[0173] Step 3: Based on the hyper - parameter combinations, and using the parameter optimization model to generate trading signals through the daily market data of the stocks in the stock pool. For example, trading signals for stocks 300502 and 300563 are generated, as shown in Tables 5 and 6 specifically;

[0174] Table 5 Example of Operation and Profit Statistics for Stock 300502 after Using Hyper - parameters

[0175]

[0176]

[0177] Table 6 Example of Operation and Profit Statistics for Stock 300563 after Using Hyper - parameters

[0178]

[0179] It should be noted that the profit calculation formulas in Tables 5 and 6 are as follows: Profit=(Selling Price - Buying Price)*Quantity;

[0180] For example: For the stock 300502, it can be seen from the data in Table 5 that trading signals are generated based on the moving average and the position - building coefficient calculated according to the closing price on September 19, 2024. On September 20, 2024, it is bought at the opening price of 94.96, and the number of shares bought is 100. Trading signals are generated based on the moving average calculated according to the closing price on November 21, 2024. On November 22, 2024, it is sold at the opening price of 118, and the number of shares sold is 50. Trading signals are generated based on the moving average and the position - closing coefficient calculated according to the closing price on November 24, 2024. On November 25, 2024, it is sold at the opening price of 113, and the number of shares sold is 50;

[0181] Therefore, the profit obtained when selling on November 25, 2024 = ((118 - 94.96) * 50)+((113 - 94.96) * 50)=2054. Since the profit is positive, it is actually a gain;

[0182] For the stock 300563, it can be seen from the data in Table 6 that trading signals are generated based on the moving average and the position - building coefficient calculated according to the closing price on September 26, 2024. On September 27, 2024, it is bought at the opening price of 32.8, and the number of shares bought is 100. Trading signals are generated based on the moving average calculated according to the closing price on October 29, 2024. On October 30, 2024, it is sold at the opening price of 58.71, and the number of shares sold is 50. Trading signals are generated based on the moving average and the position - closing coefficient calculated according to the closing price on November 3, 2024. On November 4, 2024, it is sold at the opening price of 51.82, and the number of shares sold is 50;

[0183] Therefore, the profit obtained when selling on November 4, 2024 = ((58.71 - 32.8) * 50)+((51.82 - 52.8) * 50)=2246.5. Since the profit is positive, it is actually a gain;

[0184] Output and save the information file according to the trading information. The information file is shown in Table 7;

[0185] Table 7 Summary Table of Trading Information

[0186]

[0187]

[0188] Step 4: At the end of each quarter, repeat Step 1, Step 2, and Step 3 to replace the stocks in the stock pool with new ones and update the corresponding hyper - parameter combinations to ensure that the strategy is optimized dynamically with the market;

[0189] For example, the measured annualized return rate of the ChiNext in 2024 obtained according to the above optimization steps is 22%, and the details are shown in Tables 8, 9, 10, and 11;

[0190] Table 8

[0191]

[0192]

[0193]

[0194] Table 9

[0195]

[0196]

[0197] Table 10

[0198]

[0199]

[0200] Table 11

[0201]

[0202]

[0203] From Table 8, Table 9, Table 10 and Table 11, the connection scheme between the quarterly update logic and the outstanding stocks can be obtained, forming a time window mode, keeping in sync with the market rhythm, and effectively improving the yield of the ChiNext board.

[0204] The above-described embodiments merely represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention.

Claims

1. A stock quantification method based on multi-dimensional quantitative indicators, characterized in that: The steps include: Step 1: Obtain GEM stock data, filter the stock data through multi-dimensional quantitative indicators, assign different weight coefficients to the multi-dimensional quantitative indicators, calculate the comprehensive score, and select high-scoring stocks to build a stock pool; Step 2: Construct a parameter optimization model based on a genetic algorithm, and optimize the short-term moving average parameters, long-term moving average parameters, short-term VWAP parameters, long-term VWAP parameters, and position opening and closing coefficients through the parameter optimization model to obtain a hyperparameter combination with a high backtest yield; Step 3: Based on the hyperparameter combination, the parameter optimization model is used to generate a trading signal through the daily market data of the stocks in the stock pool as a reference for the next day's trading; Step 4: Update the stock pool and the hyperparameter combination at the end of each quarter to dynamically quantify stock data.

2. A stock quantification method based on multi-dimensional quantitative indicators according to claim 1, characterized in that: The multi-dimensional quantitative indicators include: Alpha, Beta, Sharpe ratio and logarithmic cumulative rate of return; The mathematical expression of Alpha is: α=(Ri-Rf)-β(Rm-Rf) Among them, Ri is the actual rate of return of the asset; Rf is the risk-free rate of return; Rm is the market benchmark rate of return; β is the market risk coefficient of the asset; Beta mathematical expression: Among them, Cov(Ri, Rm) is the covariance between the actual return of the asset and the market benchmark return; Var(Rm) is the variance of the market benchmark rate of return; is the average of the actual returns of the assets; is the average of the market benchmark yields; t is time; Sharp ratio mathematical expression: Among them, Ri is the actual rate of return of the asset; Rf is the risk-free rate of return; σi is the standard deviation of the actual return of the asset; Mathematical expression of logarithmic cumulative rate of return: Where Pt is the asset price at time t; Pt-1 is the asset price at time t-1.

3. A stock quantification method based on multi-dimensional quantitative indicators according to claim 2, characterized in that: The comprehensive score calculation formula is as follows: Comprehensive score = weight coefficient 1*Alpha+weight coefficient 2*Sharpe ratio+weight coefficient 3*logarithmic cumulative rate of return+weight coefficient 4*(1-ABS(Beta-1)); The specific steps for calculating the comprehensive score are as follows: Read historical data of all stocks; Calculate the Alpha, Beta, Sharpe ratio and log cumulative return for each stock; Calculate the comprehensive score, and based on the comprehensive score data, filter out multiple stocks and output them as the stock pool for the current quarter.

4. The stock quantification method based on multi-dimensional quantitative indicators according to claim 1, characterized in that: A parameter optimization model is constructed based on a genetic algorithm, and the short-term moving average parameters, long-term moving average parameters, short-term VWAP parameters, long-term VWAP parameters, and position opening and closing coefficients are optimized through the parameter optimization model to obtain a hyperparameter combination with a high backtest yield. The specific optimization steps are as follows: Read the historical market data of each stock in the stock pool; The parameter optimization model constructed by genetic algorithm calculates and generates various parameter combinations; Optimize multiple parameter combinations to obtain hyperparameter combinations with high backtest yields; Output the optimization results and save them to a parameter summary file.

5. A stock quantification method based on multi-dimensional quantitative indicators according to claim 4, characterized in that: The steps to obtain a high backtest return hyperparameter combination are as follows: Set parameters; Generate trading signals based on parameters; Close positions based on trading signals; Determine the rate of return and obtain the hyperparameter combination with high backtest rate of return.

6. A stock quantification method based on multi-dimensional quantitative indicators according to claim 5, characterized in that: The setting parameters include a moving average for identifying trends, a volume-weighted average price that combines short-term VWAP and long-term VWAP to determine the direction of market capital flow, a test direction, and entry_coef and exit_coef parameter settings for dynamically adjusting entry and exit price trigger conditions; Among them, the moving average includes 3-day moving average, 5-day moving average, short-term moving average and long-term moving average.

7. A stock quantification method based on multi-dimensional quantitative indicators according to claim 6, characterized in that: Generate trading signals based on parameters, including long signals and short signals; The conditions for generating a long signal are: when the 5-day moving average ≥ the short-term moving average, the 5-day moving average ≥ the specified amplitude of the short-term moving average, the short-term VWAP > the long-term VWAP and the current position is 0, a long signal is generated; Among them, the calculation method of the specified amplitude of the short-term moving average is: short-term moving average + (ABS (short-term moving average-long-term moving average)*entry_coef); The conditions for generating a short signal are: when the 5-day moving average ≤ the short-term moving average, the 5-day moving average ≤ the specified amplitude of the short-term moving average, the short-term VWAP < the long-term VWAP and the current position is 0, a short signal is generated; Among them, the specified amplitude of the short-term moving average is calculated as: short-term moving average - (ABS (short-term moving average - long-term moving average) * entry_coef).

8. The method for stock quantification based on multi-dimensional quantitative indicators according to claim 6, characterized in that: Close positions according to trading signals, including long and short positions; Among them, long position closing includes step closing and full closing. When the 3-day moving average ≤ 5-day moving average and short-term VWAP < long-term VWAP, the first step closing is carried out, closing 50%. When the 3-day moving average ≤ 5-day moving average, the 3-day moving average ≤ the specified range of the 5-day moving average and short-term VWAP < long-term VWAP, the second step closing is carried out at this time, closing 50%. Among them, the specified range of the 5-day moving average = 5-day moving average - (ABS (5-day moving average - short-term moving average) * exit_coef); Full position closing includes: when the stop loss condition, i.e. the position loss margin is ≥ 5%, full position closing is carried out; When the target condition, i.e. short-term VWAP < long-term VWAP, close the position in full; Short position closing is divided into step closing and full closing. When the 3-day moving average ≥ 5-day moving average and short-term VWAP > long-term VWAP, the first step closing is carried out, closing 50%. When the 3-day moving average ≥ 5-day moving average, the 3-day moving average ≥ the specified range of the 5-day moving average and short-term VWAP > long-term VWAP, the second step closing is carried out, closing 50%. Among them, the specified range of the 5-day moving average = 5-day moving average + (ABS (5-day moving average - short-term moving average) * exit_coef); Full position closing includes: when the stop loss condition, i.e. the position loss margin is ≥ 5%, full position closing is carried out; When the target condition, short-term VWAP > long-term VWAP, close the position in full.

9. The method for stock quantification based on multi-dimensional quantitative indicators according to claim 1, characterized in that: Based on the hyperparameter combination, the parameter optimization model is used to generate trading signals through the daily market data of the stocks in the stock pool as a reference for the next day's trading. The specific steps are as follows: Update the market data of each stock in the stock pool; Generate trading signals through the parameter optimization model; Output information file according to transaction information and save it.

10. A stock quantification method based on multi-dimensional quantitative indicators according to claim 1, characterized in that: At the end of each quarter, the stock pool and the hyperparameter combination are updated to dynamically quantify stock data. The specific steps are as follows: Repeat steps 1, 2, and 3 to update the stock pool and hyperparameter combination of the previous quarter; Monitor open positions until they are closed; Output stock quantitative data.

Images