Quantitative investment method and device, equipment and storage medium
By using Stacking integrated learning framework, decision tree and AdaBoost algorithm to build an exponential enhancement model in a multi-factor model, the factor correlation and model complexity problems are solved, and more efficient and economical quantitative investment decisions are achieved.
Patent Information
- Application Number
- CN202411566996.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-05
- Publication Date
- 2025-02-25
AI Technical Summary
Factor correlation in existing multifactor models leads to problems such as high redundant information and model complexity and high computational cost.
Using Stacking integrated learning framework, the first layer uses multiple decision tree algorithms as the base learner, and the second layer uses the AdaBoost algorithm as the meta learner to build an exponentially enhanced model BoostedStack Tree, trains weak classifiers through multiple iterations, and adjusts sample weights. , reduce factor correlation and model complexity.
Reduce factor redundancy, reduce interference in correlation between factors, make the model pay more attention to important factors, reduce model complexity, improve computing efficiency, and reduce computing costs.
Smart Images
Figure CN119513600A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of quantitative investment technology, and in particular to a quantitative investment method, device, equipment and storage medium. Background Art
[0002] The complexity of the stock market requires investors to have higher requirements for investment and financial management. Therefore, quantitative investment has attracted widespread attention and has become an indispensable part of the financial investment system of many investors. Benefiting from the advancement of Internet technology, relevant information has been more widely disseminated and applied. As a current trend in the financial field, quantitative investment achieves more efficient and accurate investment decisions by combining models and computers.
[0003] In the field of quantitative investment, the multi-factor model is a commonly used investment strategy. The theoretical basis of the multi-factor model is to establish a regression for the selected factors based on the traditional linear method, but it also has some shortcomings, including:
[0004] (1) Factor correlation: The factors used in a multi-factor model may be correlated, which may lead to redundant information in the model and make the model more susceptible to noise. Highly correlated factors may cause the model to over-focus on certain specific signals and ignore other important signals.
[0005] (2) Model complexity: Multi-factor models usually contain a large number of factors and parameters, which increases the complexity and computational cost of the model. Managing and maintaining complex models requires more resources and technical support. Summary of the invention
[0006] The present invention provides a quantitative investment method, device, equipment and storage medium, which solve the problems existing in the existing methods, that is, redundant information appears in the model due to highly correlated factors, and the model has high complexity and calculation cost.
[0007] In a first aspect, the present invention provides a quantitative investment method, comprising the following steps:
[0008] Collect price data of multiple stocks in real time, and obtain multiple principal component factor data of different stocks based on the price data;
[0009] With the Stacking ensemble learning framework as the underlying logic, multiple decision tree algorithms are used as base learners in the first layer, and the AdaBoost algorithm is used as the meta-learner in the second layer to build an exponential enhancement model.
[0010] Multiple principal component factor data of different stocks are input into the index boosting model, multiple decision tree algorithms automatically generate distinguishing features during the decision tree construction process, and classify the multiple principal component factor data based on the distinguishing features to obtain classification results; the AdaBoost algorithm adjusts the weights of the multiple principal component factor data, and inputs them into the multiple decision tree algorithms again for classification iteration; the classification results include true positive examples and true negative examples;
[0011] Obtain the final classification results of different stocks after iteration, and select multiple stocks for buying and selling investments based on the final classification results.
[0012] Preferably, the step of obtaining a plurality of principal component factor data of different stocks according to the price data comprises the following steps:
[0013] Obtain the basic factors that affect stock prices and build a factor library based on the basic factors;
[0014] Perform validity inspection, preprocessing and single factor testing on multiple factors in the factor library to obtain multiple candidate factors; the preprocessing includes outlier removal, missing value filling, standardization and backtesting;
[0015] Principal component analysis is performed on multiple candidate factors to obtain multiple principal component factors.
[0016] Preferably, the basic factors include fundamental factors and technical factors, the basic factors include economic factors, industry factors and performance factors, the technical factors include company technology breakthrough factors and industry technology proportion factors, and the factor library includes multiple price and volume factors, multiple financial factors and multiple technical factors.
[0017] Preferably, the selecting of multiple stocks for buying and selling investments according to the final classification results comprises the following steps:
[0018] Get multiple F1 indicators of multiple final classification results;
[0019] Arrange multiple F1 indicators from high to low, and buy and sell stocks corresponding to multiple F1 indicators above the set threshold.
[0020] Preferably, before inputting the plurality of principal component factor data of different stocks into the index enhancement model, the method further includes training the index enhancement model, which specifically includes the following steps:
[0021] Collect historical data of multiple stock prices;
[0022] Obtain corresponding historical data of multiple principal component factors according to historical data;
[0023] The historical data of multiple principal component factors of different stocks are divided into training sets and test sets. The constructed index enhancement model is trained through the training set to optimize the model parameters and weights.
[0024] The trained exponential boosting model is tested using the test set.
[0025] Preferably, during the training of the index enhancement model, ST stocks and suspended stocks need to be eliminated.
[0026] Preferably, a rolling time window method is used to periodically select the latest principal component factor sample data and introduce an exponential enhancement model for training.
[0027] In a second aspect, the present invention provides a quantitative investment device, comprising:
[0028] The acquisition module is used to collect the price data of multiple stocks in real time, and obtain multiple principal component factor data of different stocks based on the price data;
[0029] The building module is used to build an exponential enhancement model using the Stacking ensemble learning framework as the underlying logic, using multiple decision tree algorithms as base learners in the first layer and the AdaBoost algorithm as a meta-learner in the second layer;
[0030] An input module is used to input multiple principal component factor data of different stocks into the index enhancement model, multiple decision tree algorithms automatically generate distinguishing features during the decision tree construction process, and classify the multiple principal component factor data based on the distinguishing features to obtain classification results; the AdaBoost algorithm adjusts the weights of the multiple principal component factor data and inputs them into the multiple decision tree algorithms again for classification iteration; the classification results include true positive examples and true negative examples;
[0031] The investment module is used to obtain the final classification results of different stocks after iteration, and select multiple stocks for buying and selling investments based on the final classification results.
[0032] In a third aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned quantitative investment method when executing the program.
[0033] In a fourth aspect, the present invention provides a computer-readable storage medium, wherein the storage medium stores a computer program, and the computer program implements the above-mentioned quantitative investment method when executed by a processor.
[0034] Compared with the prior art, the present invention has the following beneficial effects:
[0035] The present invention is based on the underlying logic of the Stacking ensemble learning framework. In the first layer, the decision tree algorithm is used as the base learner. Multiple decision tree algorithms automatically generate distinguishing features during the decision tree construction process and classify multiple principal component factor data based on the distinguishing features, giving full play to the excellent learning and feature selection capabilities of the decision tree algorithm, reducing factor redundancy, and avoiding the negative impact of highly correlated factors on the model. In the second layer, the AdaBoost algorithm with lower complexity and significantly improving model performance is used. The AdaBoost algorithm adjusts the sample weights and pays attention to multiple factors that are misclassified by the decision tree, thereby reducing the interference of correlation between factors, making the model pay more attention to important factors, and reducing the complexity of the model.
[0036] Decision trees and AdaBoost can be trained in parallel, so that the exponential boost model BoostedStack Tree can improve computational efficiency and reduce computational costs. Through multiple iterations of training weak classifiers, AdaBoost will adjust sample weights to further reduce the negative impact of related factors on the final model, thereby solving the problem of possible correlation between factors used in multi-factor models and avoiding redundant information in the model. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0038] Figure 1 A flowchart of a quantitative investment method of the present invention;
[0039] Figure 2 This is a schematic diagram of the quantitative trading backtesting results of the present invention;
[0040] Figure 3 This is a comparison chart of the results inside and outside the quantitative trading sample of the present invention. DETAILED DESCRIPTION
[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0042] The present invention provides a quantitative investment method based on a multi-factor model, with decision trees and AdaBoost algorithms as the core to construct an integrated learning index enhancement model BoostedStack Tree, so as to improve the performance and stability of the investment portfolio. The method includes collecting historical data of stock prices and the corresponding basic factors affecting stock prices. Construct a factor library, and then pre-process the factor data, test the factors, and analyze the principal components of the factors. Train the index enhancement model BoostedStack Tree, and test the index enhancement model BoostedStack Tree. Eliminate ST stocks and suspended stocks during training and testing. Collect current stock price data and stock basic factor data in real time, calculate the F1 index, and select the 20 best performing stocks based on the F1 index; allocate positions equally to execute buy operations; update the stock pool once a month; use the rolling time window method to regularly select the latest factor sample data and introduce it into the model for training; risk control, offensive profit stop, defensive stop loss. Reference Figure 1 , specifically including the following steps:
[0043] Step 1: Collect the basic factors that affect stock prices.
[0044] Collect historical data of stock prices and the basic factors that affect stock prices. Basic factors include fundamental factors and technical factors. Basic factors include economic factors, policy factors, industry factors and performance factors, while technical factors include company technological breakthrough factors and industry technology proportion factors.
[0045] Supporting or restricting the development of specific industries directly affects the stock prices of related companies. For example, restricting the development of industries will lead to adverse effects on the stock prices of related companies, such as price limits on public utility products and services, such as transportation, gas, water and electricity, which will reduce the profitability of public utility companies and lead to a drop in stock prices. The stock prices of companies that enjoy tax reduction and exemption benefits tend to rise, while raising personal income tax may lead to a decline in social consumption levels, affect commodity sales, and then affect the company's production scale and profitability, leading to a drop in stock prices. The impact of these event factors on the stock market is reflected by affecting the profitability of companies and market interest rates, thereby causing stock price fluctuations.
[0046] Industry factors can be analyzed from the perspective of products to determine whether the company's products are productive resources or consumer resources. Productive resources refer to products that meet other production needs, while consumer resources are products that are directly consumed by people. However, these two types of products are affected by the economic environment to different degrees. Generally speaking, productive resources are more affected by changes in the economic climate, while consumer resources are relatively less affected. In other words, when the economic situation improves, the demand for productive resources increases faster than that for consumer resources; conversely, when the economy is in recession, the demand for productive resources decreases faster. In consumer resources, the necessity and luxury of the product also need to be considered. Because the nature of different products will have different degrees of impact on market demand, company operations, and market prices.
[0047] By analyzing the sales targets and sales scope of the company's products through demand patterns, we can understand the company's domestic sales and export ratio. Domestic sales are mainly affected by domestic events and economic factors, while export sales are more easily affected by the international economic and trade environment. At the same time, it is necessary to investigate and analyze the degree to which the company's products meet different demand targets. Different demand targets have different requirements for product performance and quality. Therefore, the company must customize products according to demand to meet the needs of different groups. Otherwise, it will affect product sales, and then affect the company's profitability, reduce investment returns, and cause stock prices to fall.
[0048] Through production pattern analysis, it is possible to determine whether a company is labor-intensive, capital-intensive, or knowledge-technology intensive, and assign corresponding weights to each factor. Labor-intensive enterprises rely mainly on labor input, while capital-intensive enterprises rely mainly on capital input. Knowledge-technology intensive enterprises, on the other hand, pay more attention to the input of knowledge and technology. In economically underdeveloped countries or regions, the proportion of labor-intensive enterprises is usually larger, while in economically developed regions, capital-intensive enterprises are more dominant. With the continuous development of new technologies, technology-intensive enterprises have gradually replaced the position of capital-intensive enterprises. In addition, different types of enterprises have different labor productivity and competitiveness, which will also affect the sales and profitability of products, resulting in differences in investment returns.
[0049] Step 2: Construct a factor library based on basic factors.
[0050] Based on the collected data, we build a factor library, including price and volume factors (such as price-earnings ratio, price-to-book ratio, trading volume, etc.), financial factors (such as revenue growth rate, profit margin, etc.), technical factors (such as moving average, relative strength index, etc.), etc. We ensure that the factors have high information content and have a significant impact on stock prices. See Table 1 for specific factor data.
[0051] Table 1 Factor library
[0052]
[0053]
[0054] Step 3: Perform principal component analysis on multiple factors in the factor library to obtain multiple principal component factors.
[0055] Each factor in the factor library is tested for validity, preprocessed, and tested for single factors to obtain multiple candidate factors. The preprocessing process includes data cleaning, outlier removal, missing value filling, standardization, and backtesting.
[0056] Outlier removal: Remove outliers from factor data in the factor library.
[0057] Missing value filling: The missing values of the previous trading day's data are filled with the data of the next trading day in a forward traversal manner.
[0058] Standardization: Standardize the factor data to ensure the accuracy and consistency of the data.
[0059] Backtesting: Reference Figure 2 and Figure 3 , backtest the factors, use historical data to evaluate the correlation and predictive ability between each factor and stock prices, and screen out factors with better performance as the object of subsequent analysis.
[0060] Calculate the IC value of the single factor after preprocessing, that is, the specific factor value factor of the current period and the rate of return r of the next period t+1 The correlation coefficient between them is usually calculated using the Pearson correlation coefficient: IC = corr (factor t , r t+1 ).
[0061] Calculate the IR value of the pre-processed single factor to measure the stability of the factor's excess (Alpha) returns in the past: Among them, mean represents the mean and std represents the standard deviation.
[0062] When the absolute value of IR is greater than 0.1, the factor is relatively stable and multiple candidate factors are obtained.
[0063] Principal component analysis was performed on multiple candidate factors to reduce redundant information between factors and extract the main influencing factors, thereby reducing model complexity and improving prediction accuracy, and 10 principal component factors were extracted.
[0064] Step 4: Build an exponential enhancement model.
[0065] Based on the selected principal component factors, the present invention uses a decision tree algorithm as a base learner in the first layer based on the underlying logic of the Stacking ensemble learning framework on the basis of establishing a multi-factor model, giving full play to the excellent learning and feature selection capabilities of the decision tree algorithm, and uses the AdaBoost algorithm with lower complexity and which can significantly improve model performance in the second layer to reduce the overfitting degree of the decision tree algorithm, thereby constructing an exponential enhancement model BoostedStack Tree, and using the historical data of multiple stock prices to learn the complex relationships and weights between factors.
[0066] Balance the impact of related factors and adjust the weights. The decision tree automatically completes feature selection by selecting the features that can best distinguish data during the construction process, reducing factor redundancy and avoiding the negative impact of highly correlated factors on the model. The AdaBoost algorithm adjusts the sample weights and focuses on samples that are misclassified by weak classifiers, thereby reducing the interference of correlation between factors and performing automatic feature selection and dimensionality reduction.
[0067] To reduce the impact of factor noise and factor correlation, the integrated learning of multiple models under the multi-level information fusion Stacking framework helps to reduce noise and reduce the risk of overfitting: by combining multiple basic learners, the stacking method can smooth the noise and bias of a single model and reduce the noise impact caused by factor correlation.
[0068] According to the historical data of multiple stock prices, the corresponding historical data of multiple principal component factors are obtained, and the historical data of multiple principal component factors of different stocks are divided into a training set and a test set.
[0069] The constructed exponential boosting model BoostedStack Tree is trained through the training set, and the model performance is optimized by adjusting parameters and weights.
[0070] The trained exponential boosting model BoostedStack Tree is tested using the test set to evaluate its accuracy and robustness on unknown data.
[0071] Special Treatment (ST) stocks and suspended stocks during training and testing (ST stocks mean that the individual stock's operating performance has deteriorated, financial problems have occurred, or other abnormalities have occurred, and there is a high risk of delisting, which will cause investors in the market to panic and sell their stocks, causing the stock price to plummet) are excluded to ensure the reliability and accuracy of the data.
[0072] Step 5: Collect the price data of multiple stocks in real time, and obtain multiple principal component factor data of different stocks based on the price data. Input the multiple principal component factor data of different stocks into the trained index enhancement model.
[0073] Set up a real-time data interface to collect current stock prices and related fundamental factor data to update model parameters and perform real-time forecasts.
[0074] According to the F1 index obtained by training and testing the index enhancement model BoostedStack Tree, the top 20 stocks with the best performance are selected as investment targets, and the investment list is output. The calculation formula of the F1 index is as follows:
[0075] F1 = 2×TP / (total number of samples + TP-TN).
[0076] Among them, TP (True Positive) is a true positive example, and TN (True Negative) is a true negative example.
[0077] The quantitative investment system can automatically buy and sell the stock list output by the model each time, and then store the transaction data in the database, perform multi-dimensional performance verification based on the transaction results, and calculate relevant strategy returns, benchmark returns, Alpha, Beta, Sharpe ratio and maximum drawdown and other indicators.
[0078] The buy operation is executed by allocating positions equally, allocating funds to the selected stocks.
[0079] Using the rolling time window method, the latest factor sample data is regularly selected and introduced into the model for training to maintain the adaptability and effectiveness of BoostedStack Tree.
[0080] Every month, we reselect investment targets based on the latest data and update the stock pool to adapt to market changes and adjust the investment portfolio.
[0081] Perform risk control when executing buy or sell operations, set offensive take-profit and defensive stop-loss strategies. For example, when the stock price is greater than 120% of the purchase price, or the stock price falls below the lowest price in 10 days and the opening price is 95% of the purchase price, set a sell operation to clear the stock to avoid potential risks and protect investment returns, and transfer funds to asset allocation of other stocks.
[0082] The present invention integrates multiple factors: by using decision trees and AdaBoost, the model can simultaneously integrate multiple factors, including market factors, financial factors, technical factors, etc., thereby improving the model's comprehensive analysis capability of the market.
[0083] Make full use of nonlinear relationships and reduce the impact of factor correlation and model complexity: Decision Tree can handle nonlinear relationships, while AdaBoost can combine multiple weak classifiers into a strong classifier to better capture the complex relationships and patterns in the data. Decision Tree and AdaBoost can be trained in parallel, so that the exponential boost model BoostedStack Tree can improve computational efficiency and reduce computational costs.
[0084] The time window is set and the model adopts a rolling training method, regularly selecting the latest factor sample data for training. It can reflect market changes in a timely manner and adjust model parameters to adapt to the new market environment.
[0085] Risk control strategies are taken into consideration in the model design, including offensive profit-taking and defensive stop-loss, which can effectively manage investment risks and protect the safety of investors' funds.
[0086] The model can automate buying and selling operations, saving investors’ time and energy and improving investment efficiency and execution results.
[0087] Based on the same concept, the present invention also provides a quantitative investment device, including a collection module, a construction module, an input module and an investment module.
[0088] The acquisition module is used to collect the price data of multiple stocks in real time, and obtain multiple principal component factor data of different stocks based on the price data.
[0089] The building module is used to build an exponential enhancement model using the Stacking ensemble learning framework as the underlying logic, using multiple decision tree algorithms as base learners in the first layer and the AdaBoost algorithm as the meta-learner in the second layer.
[0090] The input module is used to input multiple principal component factor data of different stocks into the index enhancement model. Multiple decision tree algorithms automatically generate distinguishing features during the decision tree construction process, and classify the multiple principal component factor data based on the distinguishing features to obtain classification results; the AdaBoost algorithm adjusts the weights of the multiple principal component factor data and inputs them again into the multiple decision tree algorithms for classification iteration; the classification results include true positive examples and true negative examples.
[0091] The investment module is used to iterate the index enhancement model by adjusting the weights of multiple principal component factor data to obtain the final classification result, and select multiple stocks for buying and selling investments based on the final classification result.
[0092] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the above-mentioned quantitative investment method is implemented when the processor executes the program.
[0093] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned quantitative investment method is implemented.
[0094] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.
[0095] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A quantitative investment method, characterized in that: The following steps are involved: Collect price data of multiple stocks in real time, and obtain multiple principal component factor data of different stocks based on the price data; With the Stacking ensemble learning framework as the underlying logic, multiple decision tree algorithms are used as base learners in the first layer, and the AdaBoost algorithm is used as the meta-learner in the second layer to build an exponential enhancement model. Multiple principal component factor data of different stocks are input into the index enhancement model, and multiple decision tree algorithms automatically generate distinguishing features during the decision tree construction process, and classify the multiple principal component factor data based on the distinguishing features to obtain classification results; the AdaBoost algorithm adjusts the weights of the multiple principal component factor data and inputs them into the multiple decision tree algorithms again for classification iteration; The classification results include true positive examples and true negative examples; Obtain the final classification results of different stocks after iteration, and select multiple stocks for buying and selling investments based on the final classification results.
2. A quantitative investment method as claimed in claim 1, characterized in that: The method of obtaining multiple principal component factor data of different stocks according to price data includes the following steps: Obtain the basic factors that affect stock prices and build a factor library based on the basic factors; Perform validity inspection, preprocessing and single factor testing on multiple factors in the factor library to obtain multiple candidate factors; the preprocessing includes outlier removal, missing value filling, standardization and backtesting; Principal component analysis is performed on multiple candidate factors to obtain multiple principal component factors.
3. A quantitative investment method as claimed in claim 2, characterized in that: The basic factors include fundamental factors and technical factors. The basic factors include economic factors, industry factors and performance factors. The technical factors include company technology breakthrough factors and industry technology proportion factors. The factor library includes multiple price and volume factors, multiple financial factors and multiple technical factors.
4. A quantitative investment method as claimed in claim 1, characterized in that: The method of selecting multiple stocks for buying and selling investments according to the final classification results includes the following steps: Get multiple F1 indicators of multiple final classification results; Arrange multiple F1 indicators from high to low, and buy and sell stocks corresponding to multiple F1 indicators above the set threshold.
5. A quantitative investment method as claimed in claim 1, characterized in that: Before inputting the multiple principal component factor data of different stocks into the index enhancement model, the index enhancement model is trained, which specifically includes the following steps: Collect historical data of multiple stock prices; Obtain corresponding historical data of multiple principal component factors according to historical data; The historical data of multiple principal component factors of different stocks are divided into training sets and test sets. The constructed index enhancement model is trained through the training set to optimize the model parameters and weights. The trained exponential boosting model is tested using the test set.
6. A quantitative investment method as claimed in claim 5, characterized in that: During the training process of the index enhancement model, ST stocks and suspended stocks need to be excluded.
7. A quantitative investment method as claimed in claim 1, characterized in that: The rolling time window method is used to regularly select the latest principal component factor sample data and introduce the exponential enhancement model for training.
8. A quantitative investment device, characterized in that: include: The acquisition module is used to collect the price data of multiple stocks in real time, and obtain multiple principal component factor data of different stocks based on the price data; The building module is used to build an exponential enhancement model using the Stacking ensemble learning framework as the underlying logic, using multiple decision tree algorithms as base learners in the first layer and the AdaBoost algorithm as a meta-learner in the second layer; An input module is used to input multiple principal component factor data of different stocks into the index enhancement model, multiple decision tree algorithms automatically generate distinguishing features during the decision tree construction process, and classify the multiple principal component factor data based on the distinguishing features to obtain classification results; the AdaBoost algorithm adjusts the weights of the multiple principal component factor data and inputs them into the multiple decision tree algorithms again for classification iteration; The classification results include true positive examples and true negative examples; The investment module is used to obtain the final classification results of different stocks after iteration, and select multiple stocks for buying and selling investments based on the final classification results.
9. A computer device, characterized in that: It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the quantitative investment method described in any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the quantitative investment method described in any one of claims 1 to 7 is implemented.