Optimizing financial time series predictions with hybrid ARIMA, LSTM, and XGBoost Models

Accurate time series forecasting is crucial, particularly in stock price prediction for Microsoft (MSFT). We combine the Autoregressive Integrated Moving Average (ARIMA) model with deep learning models, specifically Long Short-Term Memory (LSTM) and XGBoost, to enhance forecasting accuracy. ARIMA captures linear trends, while LSTM and XGBoost address nonlinear patterns, providing a comprehensive analysis. All three models produced good results, but the ARIMA-LSTM hybrid model outperformed them. It achieved a Mean Squared Error (MSE) of 22.15, Root Mean Squared Error (RMSE) of 4.70, Mean Absolute Error (MAE) of 3.20, and an R-squared (R2) value of 0.99. These results highlight the potential of hybrid models, combining traditional and deep learning techniques, in improving prediction accuracy and emphasize the importance of such combinations for optimal forecasting performance.