A PREDICTIVE DATA FEATURE EXPLORATION-BASED AIR QUALITY PREDICTION APPROACH

Abstract

Rapid population growth combined with accelerated urbanization and economic expansion has significantly intensified environmental degradation in major cities across India. Among various environmental challenges, air pollution has emerged as a critical concern due to its severe implications for public health, ecological balance, and sustainable urban development. Predicting air quality levels is a complex task because pollutant concentrations exhibit strong temporal variability, nonlinearity, and dependence on multiple external factors. This study proposes an advanced air pollution prediction framework using an improved Light Gradient Boosting Machine (LightGBM) model to forecast air quality levels in selected Indian cities. The proposed system integrates historical pollution data with forecasting-based auxiliary inputs to enhance predictive capability. A sliding window strategy is employed to effectively capture high-dimensional temporal dependencies and manage large-scale time-series data. Experimental evaluation demonstrates that the enhanced LightGBM model outperforms conventional machine learning approaches in terms of prediction accuracy, stability, and robustness. The findings confirm the suitability of the proposed approach for real-time air quality forecasting and decision support in smart city and environmental monitoring applications