AirVision

AirVision: AQI Prediction & Analytics Platform

Eleven regressors, one benchmark, a live AQI dashboard.

Category
Machine Learning
  • Python
  • scikit-learn
  • XGBoost
  • LightGBM
  • CatBoost
  • Streamlit

Abstract preview - not a screenshot

01 / Problem

Predict the Air Quality Index from pollutant measurements across multiple Indian cities.

02 / Approach

  1. 01Imputation, date features, encoding, scaling, evaluation and saved artifacts.
  2. 02Benchmarked 11 regression algorithms: Random Forest, XGBoost, LightGBM, CatBoost, SVR, KNN, AdaBoost, ElasticNet, Decision Tree, Gradient Boosting, Linear Regression.
  3. 03Interactive Streamlit dashboard.

03 / Architecture

AirVision system architecturePollutant data, then Preprocessing (Imputation, Date features, Encode & scale), then 11-model benchmark, then Best: CatBoost, then Saved artifacts, then Streamlit dashboard.TRAINING PIPELINE01Pollutant dataMEASUREMENTS ACROSS INDIAN CITIES02ImputationMISSING VALUESDate featuresFROM TIMESTAMPSEncode & scaleMODEL-READY0311-model benchmarkCOMPARED ON R² · MAE · RMSERFXGBLGBMCATSVRKNNADAENETDTGBLR04Best: CatBoostR² 0.91 · MAE 12 · RMSE 1905Saved artifactsMODEL + PREPROCESSING06Streamlit dashboardLIVE AQI PREDICTION & ANALYTICS
Fig. 01 - System architectureSimplified

04 / Key details

  • AQI prediction from pollutant measurements across multiple Indian cities.
  • Benchmarked 11 regression algorithms.
  • Best: CatBoost, R² = 0.91, MAE = 12, RMSE = 19.
  • Interactive Streamlit dashboard.

05 / Results & signals

R² 0.91
CatBoost
11
Models benchmarked
MAE 12
Mean abs. error
RMSE 19
Root mean sq. error

That's every entry · back to the index

Back to work

Return to the home page, right where you left the project grid.