Gradient Boosting and LSTM Based Hybrid Ensemble Learning for Two Step Prediction of Stock Market

Abstract

Prediction of stock market price using different artificial intelligent techniques have become an efficient and effective method for stock market prediction with higher prediction accuracy.In this present work, thus we provide an ensemble technique that comprises of two base models namely extreme gradient boosting method and long short term memory method for short term prediction of stock market.Previously the prediction of stock price was confined to all the data available, irrespective of its significance in prediction accuracy.This study investigates different issues for predicting the closing price of the stock market.Based on the two step ensemble method (including a feature selection and combination of two different intelligent techniques).Convolutional Neural Network (CNN) method is used for feature selection purpose based on the correlation coefficient of different technical indicators for predicting the closing price.Additionally, ensemble learning is applied for increasing the prediction accuracy.The subset of selected input features enhances the model's accuracy.The performance evaluation of the proposed model is performed by comparing it with different other models like Support Vector Machine (SVM), Long Short Term Memory (LSTM), Kernel Extreme Learning Machine (KELM), etc.As a new addition to the previous literature the proposed combined method extracts the features that mainly influences the accuracy of the predicted price hence better result in less time is observed.The proposed ensemble learning technique exhibited the best predicted output as compared with other methods discussed in this study.