Distribution Diversity based Feature Optimization for Stock Trading Predictions using Supervised Learning

Abstract

The accurate change in forecasting stock market indices might provide individual investors and financial managers with valued information strategically. Nevertheless, estimating stock trading status remains to be a thought-provoking because of movements of stock price were characterized by maximum nonlinearity & volatility. This manuscript proposes a novel machine learning method that adopts “Distribution Diversity based Feature Optimization for Stock Trading Predictions using Supervised Learning (STPSL). Here, the ANOVA standard has used to identify the diversity between the price ticks observed for the trading inputs of status UP (positive) and DOWN (negative). Besides the selection of optimal features, the proposal adopts the classifier Adaboost to perform binary classification. The experimental study carried on the data corpus collected from the yahoo finance deliverables of stock trading ticks and the status (UP or DOWN) of the corresponding trading inputs. Performance analysis of the proposed model has carried by comparing with the other contemporary method, which lightens the significance of the proposal.