Comprehensive Analysis and Forecasting of Maximum and Minimum Temperature Using Deep Learning Techniques

Global climate change increasingly impacts weather patterns, posing challenges for agriculture, water resources, and public health. Maximum and minimum temperatures are two critical indicators of climate variability, and their accurate prediction is essential for climate adaptation strategies. This study conducts a comprehensive analysis and forecasting of maximum and minimum temperatures in Bangladesh using advanced deep learning models, including LSTM, BiLSTM, TCN, and GRU, with and without attention mechanisms. We analyzed historical maximum and minimum temperature data from 35 weather stations (1980–2023) to identify trends and patterns. The experimental results demonstrate that the Temporal Convolutional Network (TCN) with attention outperforms other models, achieving the lowest test loss (0.00256) and highest R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> (0.858), indicating superior predictive performance. These findings will enhance temperature forecasting and support Bangladesh’s climate resilience planning and disaster preparedness.