Machine Learning-Based Climate Change Modeling and Long-Term Temperature Forecasting

Climate change prediction has become a critical scientific challenge due to the complex and nonlinear interactions among atmospheric, oceanic, and terrestrial variables. Traditional physical models, although detailed, often require high computational power and exhibit limited flexibility when handling large, multi-dimensional datasets. This study presents a data-driven approach for climate change modeling using machine learning algorithms to improve long-term temperature forecasting accuracy. Historical climatic data obtained from open-access NASA and NOAA repositories were processed to train and evaluate multiple models, including Linear Regression, Random Forest, and Long Short- Term Memory (LSTM) networks. The proposed LSTM-based forecasting model effectively captures temporal dependencies and nonlinear variations in temperature data over several decades. Comparative analysis demonstrates that machine learning techniques outperform conventional regression methods in both accuracy and generalization. The developed framework offers a scalable and efficient solution for understanding future temperature trends, assisting policymakers and environmental planners in sustainability and climate adaptation strategies. The results highlight the potential of artificial intelligence as a transformative tool for climate research, enabling better prediction, monitoring, and management of climate variability.