How ram0101k/real-time-climate-disaster-predictionn works
This is a Streamlit flood-risk prediction application for Telangana districts. A user selects a district; the app fetches current weather, preprocesses it, runs prediction and visualization, then displays a generated map. A separate historical-data workflow converts or preprocesses weather data and trains the hybrid model. The prediction script also creates a confidence chart, though its displayed risk classification is threshold-based on rainfall. Runtime data files are referenced in source but are not present in the supplied tree, so they are represented through producer-to-consumer links. The README is empty; arrows reflect sampled code.
Main entry points: predict_and_visualize.py Predict and Visualize · app.py Streamlit App · train_hybrid.py Hybrid Trainer · fetch_data.py Fetch Weather
Components 13
User Interface
- Streamlit App
app.py
Weather Data
- Fetch Weather
fetch_data.py - Live Preprocessing
preprocess_live.py - Dataset Conversion
convert_kaggle_telangana_to_project_format.py - Training Preprocessing
preprocess_train.py
Modeling
- Hybrid Trainer
train_hybrid.py - Temporal Transformer
model/transformer.py - Graph Network
model/gnn.py - Model Checkpoint
model/flood_model.pth
Prediction Results
- Predict and Visualize
predict_and_visualize.py - Flood Risk Map
flood_risk_map.html
Other components
- District User
- Weather API
Connections 17
- District User → Streamlit App: selects district
- Streamlit App → Fetch Weather: runs with district
- Fetch Weather → Weather API: requests weather
- Fetch Weather → Live Preprocessing: saves weather for
- Streamlit App → Live Preprocessing: runs preprocessing
- Streamlit App → Predict and Visualize: runs prediction
- Dataset Conversion → Training Preprocessing: writes historical data for
- Training Preprocessing → Hybrid Trainer: writes training data for
- Hybrid Trainer → Temporal Transformer: calls
- Hybrid Trainer → Graph Network: calls
- Hybrid Trainer → Model Checkpoint: saves weights
- Live Preprocessing → Predict and Visualize: writes live features for
- Predict and Visualize → Model Checkpoint: loads weights
- Predict and Visualize → Temporal Transformer: calls
- Predict and Visualize → Graph Network: calls
- Predict and Visualize → Flood Risk Map: generates
- Streamlit App → Flood Risk Map: displays