Rough neuro voting system for data mining: Application to stock price prediction

This work proposes a rough neuro voting system with modified definitions of rough set approximations for knowledge discovery from complex high dimensional data. Proposed modification of rough set concepts has been used for attribute subset selection. Ensemble of neural networks are used for analysing subspaces of data in parallel and a voting system is used for final decision. The rough neuro voting system is used for stock price prediction with considering other influencing factors in addition to day-to-day stock data. The proposed approach shows effective in predicting increment or decrement of the nextday's stock price from simulation experiment.