Forecasting Significant Stock Market Price Changes Using Machine Learning: Extra Trees Classifier Leads

Abstract

Predicting stock market fluctuations is a difficult task due to its intricate and ever-changing nature. To address this challenge, we propose an approach to minimize forecasting errors by utilizing a classification-based technique, which is a widely used set of algorithms in the field of machine learning. Our study focuses on the potential effectiveness of this approach in improving stock market predictions. Specifically, we introduce a new method to predict stock returns using an Extra Trees Classifier. Technical indicators are used as inputs to train our model while the target is the percentage difference between the closing price and the closing price after 10 trading days for 120 companies from various industries. The 10-day time frame strikes a good balance between accuracy and practicality for traders, avoiding the low accuracy of short time frames and the impracticality of longer ones. The Extra Trees Classifier algorithm is ideal for stock market predictions because of its ability to handle large data sets with a high number of input features and improve model robustness by reducing overfitting. Our results show that our Extra Trees Classifier model outperforms the more traditional Random Forest method, achieving an accuracy of 86.1%. These findings suggest that our model can effectively predict significant price changes in the stock market with high precision. Overall, our study provides valuable insights into the potential of classification-based techniques in enhancing stock market predictions.