A hybrid forecasting model for prediction of stock index of tata motors using principal component analysis, support vector regression and particle swarm optimization

Abstract

This paper presents an extremely precise prediction method which improved the decision of the investors on daily direction of the stock market. Among the studies that focus on daily stock market forecasting, the hybrid machine learning techniques are more appreciated than the conventional data mining procedures. With an intent to produce such a model with more accurate predictions, this paper analyzes a series of technological indicators used in usual studies of the stock market and uses principal component analysis (PCA), along with support vector regression (SVR) and particle swarm optimization (PSO) algorithm. Feature extraction is such a procedure that can remove the unnecessary and unrelated factors, and reduce the dimension of the input variables from the original dataset. The feasibility and efficiency of the proposed PCA-SVR-PSO hybrid model was applied to forecast the daily closing prices of stock index of TATA Motors. The performance of the proposed approach is evaluated with 4304 (from 1st January 2001 to 6th April 2018) trading days historical stock price data of Tata motors collected from Bombay Stock Exchange (BSE). The total data sets were splits into two parts, 80% of the data (3444) has been used in the training phase and rest 20% of the data (860) for the testing phase. We compared our results with ANN-PSO and SVR-PSO hybrid models. The experimental results reflect that the proposed hybrid model incorporating PCA is more practicable and better performs than SVR-PSO.