ANALYSIS OF STOCK MARKET PREDICTION USING VARIOUS MACHINE LEARNING TECHNIQUES

Abstract

The objective is to compare machine learning models (SVM, Random Forest, ANN, Naive Bayes) for stock price prediction using historical data, assessing accuracy under different input methods for reliable forecasting. This study compares two methods of input data: one using ten technical characteristics (open, high, low, close prices) and the other using trend-deterministic data. Four prediction models—SVM, Random Forest, ANN, and Naive Bayes—are tested on ten years of historical data (2003-2012) from Reliance Industries, Infosys Ltd., CNX Nifty, and S&P BSE Sensex. The models' accuracy is assessed, and the efficacy of regression analysis is explored for improving forecasts. The study finds that forecasting stock price movements in 23 Indian stock markets remains challenging due to high risk and volatility. It compares two data input methods—one using technical characteristics (open, high, low, close prices) and another using trend-deterministic data—across four prediction models: Support Vector Machine (SVM), Random Forest, Artificial Neural Network (ANN), and Naive Bayes. Historical data from 2003-2012, covering Reliance Industries, Infosys Ltd., CNX Nifty, and S&P BSE Sensex, is used to assess the models' accuracy. The results indicate that traditional methods like technical and fundamental analysis are not always reliable, and regression analysis may be enhanced by incorporating more factors. ANN is highlighted as a promising approach for stock market prediction, though it requires further refinement. The study emphasizes the need for more advanced and diverse techniques from computer science and economics to improve stock price forecasting, while also recommending areas for future research in this field. The study's novelty lies in comparing advanced prediction models (SVM, Random Forest, ANN, Naive Bayes) using both technical and trend-deterministic data, highlighting the potential of ANN for improving stock forecasts.