Understanding the Dynamics of Dengue in Bangladesh: EDA, Climate Correlation & Predictive Modeling

Abstract

This thesis presents a comprehensive exploration of the intricate relationship between air quality, dengue transmission, and climate factors in Bangladesh. Comprising two interrelated projects, this research delves into critical environmental and public health challenges, offering unique insights and solutions. Air Quality Analysis: The first project dissects Air Quality Index (AQI) data from multiple Bangladeshi cities, shedding light on the often-overlooked impact of anomalies and outliers in environmental datasets. It underscores the critical role of data engineering and anomaly handling in ensuring the reliability of predictive models. Furthermore, this study evaluates the effectiveness of stringent environmental regulations, especially during the COVID-19 lockdown, in mitigating air pollution. Regional variations in AQI levels reveal vulnerabilities in major urban centers like Dhaka, Chittagong, and Gazipur, emphasizing the need for evidence-based decision-making to safeguard public health and the environment. Dengue Transmission Dynamics: The second project explores the multifaceted dynamics of dengue transmission across various divisions of Bangladesh. It not only reveals a potential link between population density and dengue outbreaks but also demonstrates how anomalies in the dataset can significantly impact model accuracy. The study highlights the robust correlation between heightened humidity levels and increased dengue cases, emphasizing the importance of targeted interventions during peak transmission periods. Additionally, this research compares Machine Learning (ML) and Deep Learning (DL) models, challenging the prevailing notion that DL models consistently outperform traditional ML methods. 5 This study stands out for its holistic approach to understanding the interplay between air quality, climate, and disease transmission. It prioritizes the meticulous handling of anomalies and outliers, highlighting their potential to skew results and impact predictive model accuracy. By drawing comparisons among different datasets and datasets with and without anomalies, this research underscores the significance of robust data pre-processing. Furthermore, the comparative analysis of various ML and DL models showcases the importance of tailored model selection and the need for a thoughtful approach to data-driven research. Our study makes a substantial contribution to environmental science, epidemiology, and data analytics. Its findings offer practical solutions and insights for policymakers, public health officials, and environmental agencies. By addressing the complexities of air quality, disease transmission, and climate variability in Bangladesh, this research paves the way for evidence-based decision-making in the face of pressing challenges, ultimately safeguarding public health and the environment.