A Stock Market Prediction Model Based on SNFS and SVM Method

In the contemporary era, machine learning (ML) for stock market predicting (SMP) has become a popular topic for both researchers and market practitioners since traditional parametric models cannot extract extremely important and useful information to predict stock price. In this study, we propose a data driven and non-parametric framework structure to achieve a more accurate prediction of the Exchange Traded Fund (ETF) trend from two angles. Firstly, this study proposes a successive nonlinear feature selection (SNFS) method, which contains multiple feature engineering techniques, to extract the important technical indicators (TIs) objectively on different time spans. Secondly, we compare two methods in order to forecast the short-term ETF trends based on the selected TIs. Several metrics including accuracy, Area Under the Receiver Operating Characteristic (ROC) Curve (AUC), confusion matrix, as well as classification report to quantify model performance. And predicted signals are evaluated by applying them to a trading strategy. Comparison shows that support vector machine (SVM) outperforms the long short-term memory (LSTM) Neural Network on our ETF dataset.