A text based Decision Tree model for stock market forecasting

Decision Tree is a well established data mining techniques equipped with several algorithms to suit both linear and non linear data set for forecasting future trend. In this paper an attempt has been made to implement text based decision tree having all discrete input variables rather than a numerical decision tree where at least one variable is need to be discrete. The available historical data and other technical indicators calculated over the numerical data set of BSE sensex and NSE nifty has been converted and normalized to textual form by certain rules and decision trees are differently constructed for BSE sensex and NSE nifty with application of C4.5 algorithm and compared with the usual decision tree generated directly by applying numerical variables for the same period. The empirical study proves the better efficacy of the proposed model by outperforming the usual decision tree.