Comparative Analysis of Stock Price Prediction Accuracy: A Machine Learning Approach with ARIMA, LSTM, And Random Forest Models

Abstract

This research investigates the comparative effectiveness of three distinct predictive models – ARIMA (Auto Regressive Integrated Moving Average), LSTM (Long Short-Term Memory), and Random Forest – in forecasting stock prices. Focusing on Tata Motors and Infosys stocks, historical data spanning a significant timeframe is collected using the finance library. These models are trained on a diverse set of features including open, close, high, and low prices to capture the underlying market dynamics. The evaluation of model performance is centred on their ability to forecast stock prices over varying prediction horizons. Metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE) are utilized to quantify the accuracy and reliability of the predictions. Through rigorous analysis, this study provides insights into the strengths and limitations of each model, offering valuable guidance to investors and market analysts. The findings underscore the significance of selecting appropriate predictive models in financial forecasting and contribute to advancing the understanding of predictive modelling techniques in stock market analysis. Additionally, the research delves into the implications of long-term dependencies in stock price forecasting, shedding light on the challenges and opportunities inherent in predicting market trends over extended periods.