Using Classification Algorithms to Predict Taiwan Stock Market - A Case Study of Taiwan Index Futures

Abstract

AbstractDue to the interaction of different factors, the characteristics of the market are constantly complex and volatile. Roughly speaking, it is somewhat difficult to accurately predict stock prices. Hence this study regards price trend prediction as a classification problem.The research purposes are to compare the effectiveness of different classification models forecast for Taiwan stock market and explore whether the use of American stock market data as features can improve the accuracy of Taiwan stock market.We use Random Forest, Logistic regression and Support Vector Machine, which is frequently used in stock price classification as models. Use closing price and trading volume of NASDAQ index, the Net Buy/Sell of Three Institutional Investors, and the trading volume of Taiwan index futures as the features. In addition to OOB Error Rate, we focused more on the measurement of model performance, and introduced six ratios extended by the confusion matrix for evaluation.The empirical results show that the nine models in this study predict the decline of Taiwan stocks more frequently than rise. But all of the models do a better job of sorting the upside than the downside. So we suggest it probably can combined with SVM III to predict the Taiwan stock market, so as to improve the prediction rate of the rise data. Through the analysis of the above model, the prediction accuracy of using the data of the US stock market combined with Taiwan stock market data is better than only using Taiwan stock market or the US stock market data.KeywordsTaiwan stock marketRandom ForestLogistic regressionSVM