Python Approach for Global Air Quality Assessment Using a Mamdani Fuzzy Inference System

Abstract

This study assesses air quality across 15 major cities worldwide using a Python-based Mamdani Fuzzy Inference System (FIS). The analysis integrates six pollutants—PM₂.₅, PM₁₀, NO₂, SO₂, CO, and O₃—from the selected global-city dataset. Each pollutant is represented through linguistic categories, and a set of fuzzy rules is applied to combine pollutant information and generate a fuzzy Air Quality Index (AQI) estimate. The aggregated fuzzy output is subsequently converted into a crisp AQI value and assigned to an air-quality category. The results show marked differences among the selected cities. Delhi, Dhaka, and Lahore record the highest fuzzy AQI values and fall within the Very Unhealthy to Hazardous range, with PM₂.₅ and PM₁₀ identified as the dominant pollutants. Beijing and Mexico City also exhibit comparatively high pollution levels. In contrast, Sydney, Toronto, Tokyo, London, and Paris show lower fuzzy AQI values and are classified within the Good to Moderate range. Bar charts and stacked pollutant visualisations are used to present overall fuzzy AQI estimates and the relative contribution of individual pollutants across cities. The findings demonstrate the applicability of a Mamdani fuzzy framework for integrating multiple air-quality parameters into an interpretable assessment and for supporting comparative evaluation of urban air-quality conditions.