Advanced Weather Forecasting with Machine Learning: Leveraging Meteorological Data for Improved Predictions

Abstract

Objectives: This study proposes a machine learning approach for improving weather forecasting accuracy with reduced resource requirements, focusing on rainfall and flooding predictions in urban regions.

Methods: The study used historical meteorological data (2009–2023) from Indian regions, applying supervised learning models, including regression and ensemble methods. Data preprocessing steps included handling missing values, outlier detection, and normalization. The models were evaluated using Accuracy, MAE and MSE value to determine prediction accuracy and reliability.

Findings: The results indicate that AI models, especially LSTM, provide significant accuracy improvements with 80.11% accuracy. These models performed well in predicting short-term weather phenomena such as rainfall in urban flood-prone areas like Mumbai. The method demonstrated the potential to produce reliable forecasts with limited computational resources. The findings complement existing research, adding value by showcasing the adaptability and scalability of resource-efficient ML models for local meteorological applications. This work highlights the practical implications for urban planning and flood preparedness.

Novelty: A cost-effective machine learning framework for accurate local weather forecasting, addressing scalability and computational constraints.