Research Based On Support Vector Machine (SVM) Of Stock Timing In Quantitative Trading

Quantitative trading is a brand-new investment method.In this domain the data is deemed to be the basement,and the model is the main core,the investment policy is the chain of thoughts.With the features of rigorous objective logic,accurate and allegro trading methods and the capability of gaining stable profit and standout risk controlling,it captures more and more attention from the research and market.Recently,with a fast domestic economic development,the securities market grows bigger and stock market and the trade are flourishing more and more.Investors not only have the opportunities unparalleled in the market,but also face the risks never appeared before.The characteristics of complexity and non-linear of stock market weaken the performance of the traditional investment analysis methods in this domain.However,the traits of quantitative trading and big data in the stock market are more and more evident and crucial.The paper selects the trading data of PingAn Bank from July 1^stt 2013 to July 10^th2018,total including 1228 trading days.38 trading features are extracted from the data,which is preprocessed by the standardization and principal component analysis.The model is constructed based on the time sequence-support vector machine theory which uses the features and then,predicts the classification of price rising or falling in the market.Grid search is applied for the model's parameter estimator.Finally,the estimator indices including Accuracy\Precision\Recall and f1-score are applied to estimate the performance of models.Backtesting is implemented after the construction,which indicates the quantitative model can gain a better performance compared to the market and possesses a good capability of risk controlling.