Rainfall Prediction Using Machine Learning

Project Overview

• Objective: Predict whether it will rain tomorrow in Australia using 10 years of historical weather data.
• Tools Used: Python (scikit-learn, pandas, matplotlib), Jupyter Notebook

Problem Statement

Traditional weather forecasting struggles with high-dimensional data. This project uses machine learning models to predict rainfall based on historical weather data collected across Australia, enabling more accurate and timely weather forecasting.

Models Used

• Decision Tree Classifier
• Random Forest Classifier
• Gaussian Naive Bayes
• Artificial Neural Network (ANN)

Data Processing

• Missing value handling using mean/mode imputation
• Label encoding of categorical features
• Train-test split with class balancing
• Performance evaluation using accuracy, precision, recall, and F1-score

Results

• Random Forest: Achieved 85% accuracy with balanced precision and recall
• Naive Bayes: High recall for rain detection, useful for safety-critical use cases
• ANN: Faced challenges due to class imbalance, requiring further tuning

Key Insights

• Strong class imbalance impacted model performance; oversampling techniques recommended
• Random Forest emerged as the most robust model with generalizable results
• Feature selection revealed "Humidity at 3pm" and "Rain Today" as strong predictors

Links

• GitHub Repo

← Back to Projects