Air Quality Forecast

A multivariate deep learning project focused on predicting PM2.5 pollution levels in Delhi using LSTM networks. The model was trained on over five years of hourly data, including key pollutants and meteorological variables. After extensive preprocessing, including imputation, normalization, and time-series feature engineering, the LSTM model was optimized with dropout and early stopping to improve generalization. The final model achieved strong performance on the test set with an R² score of 0.9746, showing its potential for use in real-world pollution monitoring and public health alerts.