| With the rapid development of artificial intelligence,quantitative investment has been accepted by domestic and foreign investors.The theoretical basis of quantitative investment lies in the ineffectiveness or weak effectiveness of the market.Computer technology is combined with certain mathematical models to practice investors’ ideas and strategies.Big data and machine learning algorithms are used to establish quantitative models to help investors build a portfolio that can beat the market.In mature foreign markets,quantitative investment is increasingly used in almost all areas of investment due to its low cost and high performance advantages.However,domestic quantitative investment is still in the development stage.Through the formulation of quantitative investment strategies to explore the ineffectiveness of the market,the investment targets are quickly and effectively selected among many listed companies,and there is still a lot of room for finding excess returns.Quantitative stock selection is an important stage of quantified investment.Through quantitative methods to discover the inherent driving factors that affect stock returns,and then using computer technology and mathematical models to select high-quality stocks,it has the characteristics of quantification and precision.In this paper,the Shanghai and Shenzhen 300 Index of the Chinese stock market is taken as the research object,and the adaptability of the support vector machine stock selection model based on multi-factors with different kernel functions in the Chinese stock market is studied.This paper comprehensively considers the financial factors and market factors that affect the dimensions of stock returns,and adds the public factors that affect stock returns-macroeconomic factors.The effectiveness of the factors is analyzed through weighted least squares regression and IC sequence tests.The effective factors are screened out;then a multi-factor kernel function support vector machine classification model is established to distinguish the similarities and differences between the third-order polynomial kernel function,seventh-order polynomial kernel function,linear kernel function,Sigmoid kernel function,and Gaussian kernel function support vector machine model for multi-factor stock selection.After training and testing,the monthly stock adjustment strategy is used to select the ten stocks with the highest probability of rising each month to build an investment portfolio,and then verify the feasibility of this investment strategy in the Chinese stock market.The following conclusions can be drawn by running the effectiveness analysis and backtesting procedures on the Wankuang Quantification Platform:the effective factors of the stock market are 36 factors such as logarithmic market value,logarithmic total assets,book leverage,return on capital;By comparing the effects of different kernel function support vector machine models for stock selection based on multiple factors,it is found that the AUC index and backtest performance of the third-order polynomial kernel function support vector machine stock selection model are generally better than other kernel functions.From the strategy backtesting net worth curve,the stock-building portfolio strategy selected by the third-order polynomial kernel support vector machine model has obtained excess returns compared with the benchmark performance portfolio strategy,and has successfully outperformed the market,proving that this strategy is used to select Stocks are effective and feasible and can provide a good reference for investors. |
