Forecasting Pollution Trends: Comparing Linear, Logistic Regression, and Neural Networks

Air pollution presents a critical environmental threat, impacting human health, biodiversity, and ecological systems worldwide. Effective forecasting of air pollution levels is essential for implementing timely interventions to mitigate these effects. While several methods for predicting air pollution exist, this study focuses on refining and comparing the efficacy of ANN, Linear Regression, and Logistic Regression. The research primarily investigates the use of multiple linear regression to analyze the influence of various predictor variables on air quality indices (AQI), coupled with logistic regression for determining likelihood functions and residual analysis. Additionally, it employs an ANN with back-propagation to enhance the accuracy of forecasts. The novelty of this research lies in the integrated application of these methods to create a robust model that leverages the strengths of each approach. The objective is to develop a predictive model that not only forecasts pollution levels more accurately but also offers insights into the relationships between multiple predictors and pollution levels. Preliminary results indicate that the proposed model improves forecast accuracy by up to 20% over traditional methods, particularly in predicting peak pollution periods. This study not only contributes to the academic field by providing a comparative analysis of predictive models but also serves as a practical tool for policymakers and environmental agencies to preemptively address air quality issues.