Transfer Learning Based Ensemble Approach for Rainfall Class Amount Prediction

Rainfall prediction is a critical component in agricultural planning, water resource management, and disaster mitigation. This project introduces a binary classification model aimed at forecasting the occurrence of rainfall using a comprehensive meteorological dataset. The dataset encompasses key features such as temperature, humidity, wind direction, atmospheric pressure, and cloud cover. To improve prediction accuracy, we employed an ensemble learning approach that integrates a Random Forest (RF) and a Multi-Layer Perceptron (MLP) classifier through a Voting Classifier mechanism. This hybrid architecture effectively combines the interpretability and robustness of RF with the non-linear learning power of MLP. Extensive data pre-processing and feature engineering techniques were applied to refine input quality and enhance model performance. Experimental results indicate that the ensemble model delivers superior accuracy and generalization capability, underlining its potential utility in real-world weather forecasting applications.

Keywords:
Rainfall Prediction, Ensemble Learning, Voting Classifier, Random Forest, Multi-Layer Perceptron, Binary Classification, Meteorological Data, Weather Forecasting, Machine Learning.