rahularya3749/rainfall-classification-nasa-gujarat
Rainfall prediction for Gujarat, India (2011–2024) using weather data from NASA POWER API and machine learning models.
rainfall-classification-nasa-gujarat
Rainfall prediction for Gujarat, India (2011–2024) using weather data from NASA POWER API and machine learning models.
☔ Rainfall Prediction in Gujarat, India (2011–2024)
📊 Interactive dashboard built in Power BI showcasing rainfall trends, humidity comparison, and rainfall category distribution.
This project builds a Machine Learning model to predict daily rainfall occurrence (rain = yes/no) using historical weather data sourced from NASA’s POWER API. It demonstrates a complete end-to-end data science workflow, from API data collection to feature engineering, model building, evaluation, and interpretation.
📦 Dataset Details
- 📡 Source: NASA POWER API
- 📍 Location: Gujarat, India
- 🧭 Coordinates: Latitude: 23.223, Longitude: 72.650
- 📆 Timeline: January 2011 – December 2024 (5114 days)
- 🎯 Task: Binary Classification (Rain = Yes or No)
🔍 Project Highlights
✅ Objectives
- Fetch daily weather data from NASA’s API
- Conduct Exploratory Data Analysis (EDA)
- Perform feature engineering and selection
- Train and evaluate classification models
- Fine-tune hyperparameters to maximize performance
📊 Key Features Used
| Feature | Description |
|---|---|
T2M |
Average temperature (°C) |
T2M_MAX |
Maximum temperature (°C) |
T2M_MIN |
Minimum temperature (°C) |
RH2M |
Relative humidity (%) |
WS2M |
Wind speed at 2 meters (m/s) |
WD2M |
Wind direction at 2 meters (°) |
PS |
Surface pressure (kPa) |
PRECTOTCORR |
Precipitation (mm/day) |
day |
Date (in YYYYMMDD format) |
⚙️ Techniques Used
- 📡 API Integration – Real-time NASA POWER API
- 📊 EDA – Histograms, boxplots, KDE plots, wind rose
- 🧠 Machine Learning – Logistic Regression, SVM, Decision Tree, KNN
- 🧪 Feature Engineering – Direction binning, seasonal features
- 🧹 Feature Scaling & Selection – StandardScaler, RFE
- 🧮 Model Evaluation – AUC-ROC, precision, recall, F1-score
- 🔧 Hyperparameter Tuning – GridSearchCV (cross-validation)
📈 Model Performance
| Metric | Value |
|---|---|
| ROC-AUC Score | 0.9609 ✅ |
| Accuracy | 90.0% |
| Precision (Dry) | 0.94 |
| Recall (Rain) | 0.91 |
| F1 Score (Rain) | 0.87 |
The model shows excellent discrimination ability, strong recall for rainy days, and reliable precision for dry days. It is well-suited for real-world rainfall prediction tasks.
📁 Project Structure
rainfall-classification-nasa-gujarat/
├── 00_README.md # ✅ Overview, insights, setup
├── 01_rainfall_prediction_final.ipynb # 📓 Complete pipeline: Fetch → EDA → Modeling
├── 02_Gujarat_Rainfall_Dashboard.pdf # 📊 PDF snapshot of Power BI dashboard
├── 03_Gujarat Rainfall Dashboard.pbix # 🧩 Power BI source file (editable)
├── dashboard_preview.png # 📷 Dashboard image preview for README
🌦️ Real-World Applications
- Agriculture – Predict rainfall for crop irrigation and sowing schedules
- Disaster Management – Early warnings for heavy rainfall events
- Water Resource Planning – Monitor seasonal rainfall patterns
👤 Author
Rahul Arya
Aspiring Data Scientist | B.Sc. Physics | IBM & Stanford ML Certified
🔗 LinkedIn • GitHub
✅ Acknowledgements
- NASA POWER API – for freely accessible, high-quality climate data
- IBM Data Science Capstone – for the project framework and structure
- scikit-learn, pandas, seaborn – for modeling and visualization
📌 Note
This notebook highlights a real-world ML workflow with a high-performing model ready for deployment or further research. Ideal for portfolios targeting roles in Data Science, Climate Analytics, or AI for Social Good.