A Comparative Analysis of CAPM, ARIMA and Random Forest Models in Stock Return Prediction

Predicting stock returns is a key task in finance. Accurate forecasts help investors make better decisions, reduce risks, and improve portfolio performance. Many models have been developed for this purpose, but differ in methods, assumptions, and use cases. This paper compares three widely used models for stock return prediction: the Capital Asset Pricing Model (CAPM), the AutoRegressive Integrated Moving Average (ARIMA) model, and the Random Forest machine learning model. CAPM is grounded in financial theory and is commonly used in asset pricing. The ARIMA model focuses on time-series analysis and short-term forecasting. The Random Forest model uses machine learning to find patterns in large datasets. The models are compared based on their logic, input needs, accuracy, and interpretability. Findings show that each model has strengths and weaknesses. CAPM is simple but limited in scope. ARIMA works well for trend-based data but struggles with nonlinear factors. The Random Forest model provides high accuracy but requires more data and computing power. The best choice depends on the user's goals and available resources.