MPD A METEROGOLICAL AND POLLUTION DATASET A COMPREHENSIVE STUDY OF MACHINE AND DEEP LEARNING METHODS FOR AIR POLLUTION FORECASTING

Abstract

This study presents a comprehensive framework for air quality prediction using a Meteorological and Pollution Dataset (MPD) by integrating both machine learning and deep learning approaches. The dataset consists of pollutant concentrations (PM2.5, PM10, NO, NO₂, CO, SO₂, O₃) along with meteorological parameters such as temperature, humidity, and wind speed collected from multiple Indian cities. Data preprocessing techniques including missing value imputation, normalization, and feature engineering are applied to enhance data quality. The proposed model employs the XGBoost algorithm for AQI classification and a Long Short-Term Memory (LSTM) network for time-series forecasting of AQI values. The hybrid approach leverages the strength of XGBoost in handling structured data and LSTM in capturing temporal dependencies. Experimental results demonstrate that the model achieves high prediction performance, with classification accuracy reaching approximately 92– 95% and low forecasting error (MSE < 0.1). The findings indicate that integrating meteorological and pollution data significantly improves prediction accuracy and reliability, making the proposed framework suitable for real-time air quality monitoring and decision-making systems.