Investment Strategy Of Multi-Factor Quantitative Stock Selection Based On Boosting

Quantitative investment is a new type of investment method.It uses advanced computer technology to mine potentially effective information from a large amount of data,and quickly tracks market changes to discover new profit opportunities,while avoiding mood swings and experiential bias.The impact of gender,resulting in a more effective and easy to implement strategic plan.Although China's quantitative investment started late,with the continuous improvement of financial policies in recent years,it has provided a better investment and financing environment for investment institutions and investors.Today,China's quantitative funds have begun to take shape.Quantitative investment has embarked on the path of rapid development,attracting more and more scholars to engage in research on quantitative investment methods and investment strategies.Therefore,it is of great theoretical and practical significance to apply excellent algorithm models to the A-share market to construct quantitative strategies.As one of the most applied and researched strategies in China's quantitative investment field,the multi-factor stock selection strategy has the advantages of high stability and large capital capacity.The selection of factors and the construction of classification models are two core modules of multi-factor stock selection strategies.For the selection of factors,this paper uses the five major categories of scale,valuation,growth,profit,and leverage to directly obtain or derive sixty-two specific factors,compared with the traditional multi-factor model in terms of quantity and type.There are certain expansions.The contribution of factors to the model is output,and the effectiveness and importance of each factor are analyzed.For the construction of classification models,this paper innovatively combines the Boosting model with Logistic regression to improve the generalization ability and robustness of a single model.Using the Boosting model to construct a feature combination of the candidate factor library to replace the subjective selection process,the stock market is a complex non-linear system with a low signal-to-noise ratio.This method can improve the multifactor model's ability to obtain excess returns on stocks.,And then use the output feature combination as an input value to the Logistic model for classification.The Boosting?Logistic model theoretically proposes a new scheme for the multi-factor model.By backtesting the actual investment strategy based on the Boosting?Logistic model built on XGBoost and Light GBM,the strategy in this paper has achieved excellent performance in terms of profitability and risk control.In terms of classification performance,the ACC index of the Boosting?Logistic model is slightly better than the original Boosting model and other ensemble learning(random forest and GBDT)models,but the ACC index is basically maintained at the same level as a whole.In terms of profitability,the accumulation of the Boosting?Logistic model The return rate is much higher than the original model and other models,and the average annualized return rate has also increased by 15% to 23%.At the same time,in terms of risk,the overall volatility is lower,and the Sharpe ratio is higher overall.The maximum backtest is close to other models This model also excels in anti-risk.Among them,the best performing Light GBM?Logistic model has a cumulative annual return rate of324.01%,an average annualized return rate of 34.45%,an annualized return rate of up to 135.05%(2015),and a Sharpe rate of 0.87.They have achieved excellent performance in risk control.