Improved Decision Tree Model for Prediction in Equity Market Using Heterogeneous Data

Abstract

The popularity of decision tree forest is due to its superior performance and accuracy. The accuracy of the decision tree forest algorithm depends upon the base learner and its diversification. As far as the literature is concerned, a large number of academia’s and researchers proposed various methods which are mostly based on pre-filtering and post-filtering of the decision tree forest technique. In this research work, a novel technique is proposed for increasing the mixture of individual decision tree present in the forest, which will improvise the final precision. In the proposed method, throughout the training, each tree of the forest is being trained to use dissimilar sets of rotation spaces which are linked together to an elevated space at the parent node. After linking each rotation space, the search for the good split is done within the elevated space. The decision of selecting a rotation technique for all the succeeding nodes depends upon the placement of a good split. Conventional equity market forecasting methods are mostly based on historical data and it is used to predict the upcoming pattern. As internet information is growing in an exponential manner, few authors have proposed work based on financial news and technical indicators for improving the prediction on the equity market. In this research work, heterogeneous information from various sources like social media, world market performance, global news, financial news and historical data have been considered for improving the prediction of Indian market indices. The performance of the proposed technique is evaluated on the Indian stock market indices with significant accuracy.