Research On The Key Technology Of Securities Investment Quantitative Strategy

In recent years,artificial intelligence technology has developed rapidly.The securities investment industry,as the most efficient industry in the economy and society,has always attached great importance to the application of artificial intelligence in the financial field.In the field of securities investment,quantitative investment is gradually emerging,which refers to the stock market trading behavior when people predict stocks and issue trading instructions through computer program to realize reliable quantitative investment strategies and make investors profit.Therefore,the quantitative strategy of applying machine learning related models and algorithms to securities investment targets has also become a research hotspot in the academic circle and industry.In quantitative research strategy,stock selection strategy and timing trading strategy are important links in investment strategy formulation.Aiming at the stock selection strategy and timing trading strategy,this paper gives the research demonstration of securities investment quantitative strategy based on artificial intelligence,and provides the technical reference for the formulation of securities investment quantitative strategy.The research contents of this paper mainly include:(1)Quantitative stock selection research based on recursive feature elimination method and Stacking integrated learning.In this study,the CSI 300 is the investment target pool,and selected24 secondary indexes of fluctuation index,income index,classical technical index and trading index as the data set.After preprocessing the data set,the data set was screened by using the recursive feature elimination method,Finally,14 technical indicators were selected as the input of the model,and the top 20% of stocks of CSI 300 were marked as category 1,and the bottom20% of stocks of CSI 300 were marked as category 0,At the same time to build Stacking integrated learning,the traditional random forest,support vector machine and logical regression four machine learning algorithm classification model experimental comparative analysis,prediction selection investment target pool weekly frequency yield in the top 20% of the stock mark,found based on recursive feature elimination method and Stacking integrated learning model of the best prediction performance,Its AUC value reaches 0.7982,the accuracy rate is72.35%,the precision rate is 72.81%,the recall rate is 71.63%,and the F1 value is 72.21%,It is better than the traditional single model of random forest,support vector machine and logical regression,providing investors with referable investment strategies.(2)Quantitative timing research based on BiLSTM + Attention.In this study,taking the CSI300 index as the target,study the stock trading points,select 15 effective technical indicators as model input.After pre-processing the data,BiLSTM + Attention model is established to predict the stock buying and selling points,If the yield of the forecast stock on the day is greater than the yield of the day before yesterday,if the yield of the forecast stock on the day is less than the yield of the day before yesterday,the position is sold,At the same time,build average timing,RSRS timing,RNN timing,BiLSTM timing,and compare with them,the CSI 300 index yield as the benchmark yield,confirmed the BiLSTM + Attention quantitative timing yield is good,annualized yield of 14.97%,sharpe ratio of 0.788,maximum retracement of 9.68%,benchmark yield is-5.20%,more than the benchmark yield,and better than the other five alternative strategies,to provide certain reference for investors.Aiming at the stock selection strategy and timing trading strategy,the calculation of artificial intelligence can make the quantitative strategy research achieve a high effect,and provide reference and support for many investors in the securities market to get excess returns.