Application of Machine Learning Techniques for Stock Price Prediction

Abstract

This study highlights the robustness of selected Machine Learning models in Stock Closing Price prediction, which effectively mitigates several defects and hazards resulting from its indeterminable nature. From different literature consulted, so many tools and techniques have been used widely to tackle the prediction of the stock market In this paper, our focus is to develop a machine learning predictive model for Stock Closing Price using three different machine learning (ML) algorithms. The data inputted into the predictive model was obtained from the Yahoo Finance online repository. The dataset used contains five hundred and eighty-eight instances. Neural Networks (ANN), SVM, and KStar Algorithm were employed for the stock market price predictive models. Stimulated on Waikato Software for Knowledge Analysis (WEKA). Comparative analysis for Stock Closing Price and predictions were performed using MAE, RMSE & RAE. The results of the predictive models revealed that the correlation coefficient of the model developed for SVM, ANN, and KStar Algorithm have values of 0.0104, 0.7184, and 0.4819 respectively. So also, the Root Relative Squared Error of 107.6185%, 96.8064%, and 88.71% for SVM, ANN, and KStar respectively. This shows that ANN has a strong positive relationship frame followed by SVM and KStar. The KStar shows a faster learning time of the three algorithms followed by SVM and ANN. The study concluded that Artificial Neural Network (ANN) performs better than SVM and KStar in predicting Stock Market Prices.