Air Quality Index (AQI) Prediction in Holy Makkah Based on Machine Learning Methods

Abstract

Makkah draws millions of visitors during Hajj and Ramadan, establishing itself as one of Saudi Arabia’s most bustling cities. The imperative lies in maintaining pristine air quality and comprehending diverse air pollutants to effectively manage and model air pollution. Given the capricious and variably spatiotemporal nature of pollution, predicting air quality emerges as a notably intricate endeavor. In this study, we confronted this challenge head-on by harnessing sophisticated machine learning techniques, encompassing the fine decision tree (FDT), ensemble boosted tree (EBOT), and ensemble bagged tree (EBAT). These advanced methodologies were enlisted to project air quality index (AQI) levels, focusing specifically on the Makkah region. Constructed and trained on air quality data spanning 2016 to 2018, our forecast models unearthed noteworthy insights. The outcomes revealed that EBOT exhibited unparalleled accuracy at 97.4%, astutely predicting 75 out of 77 samples. On the other hand, FDT and EBAT achieved accuracies of 96.1% and 94.8%, respectively. Consequently, the EBOT model emerges as the epitome of reliability, showcasing its prowess in forecasting the air quality index. We believe that the insights garnered from this research possess universal applicability, extending their potential to regions worldwide.