Hybrid Ensemble Learning for Stock Market Prediction Using Time Series

Abstract

Every stock market investor is concerned with the right price to enter or exit the market, so as to maximize their returns. However, stock price prediction is a daunting task, as the stock markets are very volatile and even illogical. The machine learning algorithms are capable of finding hidden structures within the data and predict how they will affect them in the future. This paper has implemented the ARIMA model comprising 10 years' stock prices of 25 different stocks. The prediction accuracy has been measured using Ensemble Learning. In Ensemble Learning, the power of different machine learning models is combined to find the accuracy. For the proposed work, five different learning models such as Logistic Regression Model, Decision Tree Model, Support Vector Machine Model. K-Nearest Neighbor Model and Naïve Bayes Model are taken. The results of these algorithms are applied to Max Voting Classifier method for selecting the algorithm which gives the highest accuracy. The accuracy which has been predicted mostly by the weak learners will be the final prediction of the Ensemble Model. The hybrid ensemble model can outperform individual models because the accuracy is verified by observing the confusion matrices and cross validation score of each individual model.