Lahore's Air Quality Revealed: A Data-Driven Analysis Using Supervised Learning

Abstract

Ensuring safe and clean air quality for breathing is crucial as it is getting effected by several factors. This study compares data taken at midnight and afternoon to find the temporal differences and possible contributing factors to examine the Air quality in Lahore. Air quality indices (AQI) and environmental variables are modelled using machine learning (ML) models (i.e. Linear regression, decision tree, random forest, SVM, KNN, Gradient boosting, Lasso regression, Ridge regression.) to study pattern variations in the quality of the air between the two periods, which is indicative of the impact of atmospheric dynamics. Classification is being done on the data using machine learning models which have been evaluated based on precision, accuracy, recall etc. Comparison of both afternoon and midnight data have been studied and by the result we have made the conclusion that at midnight AQI is worse than the morning AQI. Decision-makers can use these findings to improve public health and protect the environment. In both periods, PM₂. ₅ and PM₁₀ turn out to be the most favorable predictors of AQI for both periods (Afternoon: Nighttime: PM10 B = 45.30, p <.05; Midnight: PM10 B = –86.2, p <.05, whereas NO₂ has a negative association in the afternoon model (B = –29.48, p <.05). There is a significant difference between mean AQI at midnight (M=223.9) and afternoon (M=194.5), which means that nocturnal air quality is poorer. In regression techniques, Lasso is the best performing method during afternoon (minimum MSE= 1567.2, maximum R2 = 0.24 for classification tasks). Simple linear regression works best at midnight, decision tree and random forest classifiers are better than SVM and KNN in attaining a perfect accuracy (100%) or classification of AQI levels (“Poor”, “Unhealthy”, “Severe”) at 00:00 hours and 12:00 hours. Based on these results, it can be seen how Lahore air pollution changes throughout the day and how PM becomes of the highest significance for the overall AQI.