How sivaprakash2k3/aqi-prediction-project works
This is a single-script AQI forecasting project: a user runs the script, which loads pollution history, prepares features, trains and evaluates tabular and sequence models, plots predictions, and saves selected models. The README documents all model variants and outputs; the excerpt shows representative preprocessing, model training, evaluation, plotting, and saving. The script is presented as a Colab workflow using Google Drive. Gaps in the excerpt limit evidence for some individual model links, so documented capabilities are included without asserting unshown implementation details.
Main entry points: aqi_prediction_project.py Forecast metrics
Components 18
Data preparation
- Load dataset
aqi_prediction_project.py - Clean and features
aqi_prediction_project.py - Tabular split and scale
aqi_prediction_project.py - 24-hour sequences
aqi_prediction_project.py
Machine learning
- Random Forest
aqi_prediction_project.py - XGBoost
aqi_prediction_project.py - LightGBM
aqi_prediction_project.py - CatBoost
aqi_prediction_project.py
Deep learning
- LSTM
aqi_prediction_project.py - GRU
aqi_prediction_project.py - Stacked GRU
aqi_prediction_project.py - CNN + LSTM
aqi_prediction_project.py
Evaluation and outputs
- Forecast metrics
aqi_prediction_project.py - Actual vs predicted plots
aqi_prediction_project.py - Saved models and scaler
aqi_prediction_project.py
Other components
- Project user
- Google Drive
- Historical AQI data
Connections 29
- Project user → Load dataset: runs
README.md - Google Drive → Load dataset: provides dataset
aqi_prediction_project.py - Historical AQI data → Load dataset: read from
aqi_prediction_project.py - Load dataset → Clean and features: prepares data
aqi_prediction_project.py - Clean and features → Tabular split and scale: scales and splits
aqi_prediction_project.py - Clean and features → 24-hour sequences: builds sequences
aqi_prediction_project.py - Tabular split and scale → Random Forest: trains and predicts
aqi_prediction_project.py - Tabular split and scale → XGBoost: trains and predicts
aqi_prediction_project.py - Tabular split and scale → LightGBM: trains and predicts
aqi_prediction_project.py - Tabular split and scale → CatBoost: trains and predicts
aqi_prediction_project.py - 24-hour sequences → LSTM: trains and predicts
README.md - 24-hour sequences → GRU: trains and predicts
aqi_prediction_project.py - 24-hour sequences → Stacked GRU: trains and predicts
aqi_prediction_project.py - 24-hour sequences → CNN + LSTM: trains and predicts
aqi_prediction_project.py - Random Forest → Forecast metrics: evaluated by
aqi_prediction_project.py - XGBoost → Forecast metrics: evaluated by
README.md - LightGBM → Forecast metrics: evaluated by
aqi_prediction_project.py - CatBoost → Forecast metrics: evaluated by
aqi_prediction_project.py - GRU → Forecast metrics: evaluated by
aqi_prediction_project.py - LSTM → Forecast metrics: evaluated by
aqi_prediction_project.py - Stacked GRU → Forecast metrics: evaluated by
README.md - CNN + LSTM → Forecast metrics: evaluated by
README.md - Forecast metrics → Actual vs predicted plots: plots predictions
aqi_prediction_project.py - GRU → Saved models and scaler: saves model
aqi_prediction_project.py - LSTM → Saved models and scaler: saves model
aqi_prediction_project.py - Stacked GRU → Saved models and scaler: saves model
aqi_prediction_project.py - CNN + LSTM → Saved models and scaler: saves model
aqi_prediction_project.py - CatBoost → Saved models and scaler: saves model
README.md - Google Drive → Saved models and scaler: stores artifacts
aqi_prediction_project.py