Deep Learning vs Machine Learning for Dry Spell Modeling
How do we model the patterns of drought? This project explores how well Machine Learning and Deep Learning methods can capture the behavior of dry spell sequences—periods without rainfall critical to agriculture and water management.
By comparing different modeling approaches, I aim to assess their performance, interpretability, and potential use for real-world forecasting in climate-sensitive regions.
Project Highlights
- 🌍 Applied Climate Modeling: Focuses on reproducing dry spells based on historical weather and satellite data.
- 🤖 ML vs DL: Benchmarks classical Machine Learning models (like Random Forests) against Deep Learning architectures (like LSTMs).
Tech Stack
- Data Processing: Pandas, NumPy
- Machine Learning: Scikit-learn (Random Forest, SVM)
- Deep Learning: TensorFlow/Keras (LSTM), PyTorch
- Visualization: Matplotlib + Seaborn
Key Insights
- Deep Learning models better capture temporal dependencies but require more data and tuning.
- ML models offer faster training and easier interpretability—making them viable for some real-world applications.
For the full code, visit the GitHub repository