The Research Of Stock’s Turning Point Based On SMOTE-SVM

With the development of economy and the change of financing concept,under the guidance of relevant policies,China’s stock market is gradually developed.In the western mature securities market,discovering useful information in financial data has important significance for investment and financing,risk assessment,industry analysis,and business management.Quantitative investment takes advantage of advanced information technology to mine the information in historical data rapidly and efficiently,which overcomes the problem of subjectivity in trading strategy,also relaxes the speculation on the decision.In recent years,quantitative investment in China is increasingly hot.For technical analysis in the past,researchers are most concerned about predicting the trend of stock price.From the capital asset pricing theory at first,to the regression model,ARIMA method,GARCH method and other methods,then the application of intelligent method such as artificial neural network,decision tree,various methods provide a valuable reference for the financial system to analyze of the market.However,because the stock market is a perplexing system,there are so many factors such as macro environment,investment psychology,business philosophy,human operation,emergency that determine the price of stock.It is difficult to obtain satisfactory results in forecasting the stock price trend.Comparing to the specific price,stock investors always concerned with the fluctuation of stock price,therefore researching the reversal point plays particularly important role for analysis.In order to effectively identify the stock price reversal point,combined with the technical analysis indicators and machine learning algorithms,this paper build a set of reversal point mining mechanism.Firstly,I selected Shenzhen real estate stock price in 2014-2016 as samples,calculated of technical indicators such as RSI,MACD,KDJ,BOLL to construct the feature engineering;then,calculated a true reversal point vector,to obtain two sets including upward reversal point data and downward reversal point data,and used SMOTE based on the imbalance problem;then,build support vector machine prediction model with evaluating and optimizing the parameters;finally,defined trading rules for back-test to confirm the validity of this model.The results show that: compared to using the original data,pre-processing the imbalance sample has higher accuracy;according to the prediction results of inversion point recognition model,the investment rate is higher than the stock’s real return rate;using a plurality of technical indicators shows better prediction effect than single index.