Quantitative Investment Strategy Based On SVM And BP Neural Network

In recent years,with the gradual completion of China's financial market,more and more financial investors have entered the A-share market.In addition to traditional fundamental investments,quantitative investment transactions have also begun to emerge.Quantitative investment traders have been unable to achieve arbitrage in the market well using traditional technical indicators.Therefore,more and more quantitative investment traders have begun to use machine learning methods to capture arbitrage opportunities.This article uses two algorithms,SVM and BP neural network,to predict the rise and fall of the line relative to the closing price of the day every 5 minutes,so as to form a certain arbitrage opportunity.The training process uniformly selects the method of rolling training set data,that is,predicts and classifies the future 1,000 data for every 24,000 data.The classification method is three categories,corresponding to the three categories of "up","down" and "flat",and the feature is selected as K Nine characteristics of line price change rate,transaction volume change rate,volatility rate,etc.,are processed using a new data standardization method.When selecting parameters,genetic algorithm tuning is introduced into the SVM algorithm to adjust the parameters of the SVM.By comparing the ROC curve,recall rate and F1 value of the two algorithms of SVM and BP neural network,it is found that the SVM model is better than the BP neural network for processing stock data sample sets.Judging from the trading results,the support vector machine algorithm model 2020 CSI 300,CSI 500 and SSE 50 indices will have a comprehensive annualized return of 70.02%,with a maximum drawdown of-3.22%,and a Sharpe ratio of 3.79.The comprehensive annualized return of the three indexes of BP neural network is 51.65%,the maximum drawdown is only-6.3%,and the Sharpe ratio is 2.37.Through the comparison of the backtest results,the strategy with better returns is also the support vector machine model.Therefore,in the subsequent actual operations,the support vector machine model is preferred for simulated trading.This article summarizes the complete process of quantitative investment research from data cleaning to data back-testing,and the benefits are significant.It has certain guiding significance for investment traders and quantitative investment beginners.