Comparative Analysis of Linear Regression, Polynomial Regression, and ARIMA Model for Short-term Stock Price Forecasting

Abstract

This research investigates the effectiveness of three prominent stock price prediction methodologies: Linear Regression, Polynomial Regression, and AutoRegressive Integrated Moving Average (ARIMA) model. The study leverages one and a half years of historical data from Apple, Tesla, Amazon, and Nike stocks to predict average prices over the ensuing 14 days. The predictive efficacy of each model is tested against actual data, revealing their respective strengths and limitations. Linear Regression offers an overview of stock trends with limited intricacy, while Polynomial Regression delivers a more comprehensive understanding of price variations and cyclical trends. However, Polynomial Regression's reliability for predictions remains uncertain. In contrast, the ARIMA model, designed explicitly for short-term forecasting, demonstrates superior accuracy, correctly predicting seven out of eight scenarios. It should be noted that this is despite its assumptions of linearity and stationarity. The findings underscore the complexity of accurate stock market prediction and highlight the ARIMA model's reliability for short-term forecasts. Therefore, understanding the strengths and weaknesses of each model is crucial for improving stock price prediction techniques and for better grasping the complex dynamics of volatile stock markets.