Market-Based Stock Allocation Using a Hybrid Regression Model

Sanjay Chakraborty, Animesh Kairi, Nilanjana Dutta Roy, Saikat Maity, Saroj Kumar Pandey, Indrajit De

Abstract: Regression analysis is intended to assist investors in identifying practical trends from historical data that aid in the formulation of their investment decisions. Stock market prediction is a technique for estimating future stock and other financial value prices for a corporation. Regression analysis is one of the most effective tools for predicting the stocks and market conditions in the stock market. The volume and velocity of information produced by the stock market are truly staggering. An effort at predicting the direction of the stock market is attempted in this paper. In this work, we have provided a comprehensive review of the most popular efficient regression methods for forecasting daily future stock prices using historical data. For short-term investment, daily prices prediction plays an important role. A mixed kind of linear and non-linear regression (Back-propagation) algorithm is proposed in this work to forecast the opening price of any company's stock. A comparison among linear, non-linear and mixed-type regression models on four widely used stock datasets are discussed and analyzed in terms of RMSE, MAE and prediction accuracy. The proposed mixed-type regression technique provides ~ 99% average prediction accuracy for the opening price.