Comparative Analysis of AutoGluon and Bayesian Optimized LSTM for Long-Term Temperature Forecasting

In this paper, two models for long-term temperature forecasting-AutoGluon and Bayesian Optimized Long Short-Term Memory-are compared. Because reliable temperature forecasts are essential for industries like agriculture and disaster preparedness, this study compares the effectiveness of an automated machine learning (AutoML) framework called AutoGluon versus a Bayesian tuned LSTM model that was tuned using Bayesian techniques. The evaluation focuses on key metrics including Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and $R^{2}$ values. The results show that the Bayesian Optimized LSTM performs better than AutoGluon on all criteria, especially when it comes to handling significant temperature fluctuations. It additionally indicates improved accuracy and resilience. The Bayesian Optimized LSTM is perfect for applications needing high precision, despite its longer training times and greater processing requirements. By comparison, AutoGluon offers a more expeditious and user-friendly solution, but its accuracy is inferior to that of the Bayesian Optimized LSTM. The findings imply that the Bayesian Optimized LSTM performs better in high-precision tasks, while AutoGluon is appropriate for cases that prioritize ease of use and quick outcomes.