APPLICATION OF REGRESSION ALGORITHMS FOR PREDICTIVE ANALYSIS IN TABRIZ

Abstract

In recent times, there has been a significant surge in the global awareness of environmental issues, prompting substantial investments by various countries and international bodies to enhance data acquisition, processing capabilities, and information retrieval systems. This heightened focus aims to prevent environmental crises that pose threats to both natural ecosystems and human well-being. Among the foremost environmental concerns under scrutiny is air pollution, exacerbated by the burgeoning global population and the proliferation of pollution sources. This academic paper utilized a dataset containing air pollution data from Tabriz spanning the years 2017 to 2019. Furthermore, a diverse array of Machine Learning algorithms was deployed to predict PM2.5 concentration levels, including Linear Regression, Lasso Regression, Polynomial Regression, Decision Tree Regression, Random Forest Regression, and XGBoost Regression. After applying these algorithms, their respective results were meticulously compared to identify the most optimal model. The XGBoost Regression algorithm emerged as the most effective, boasting an accuracy rate of 84.046%. In contrast, the Lasso Regression algorithm demonstrated the least productivity, yielding an accuracy rate of 37.101%. This thorough comparison facilitated the selection of the XGBoost Regression algorithm as the optimal choice for predicting PM2.5 concentration levels in the air, offering valuable insights for environmental monitoring and management endeavors. Keywords: air pollution, machine learning, linear regression, random forest, xgboost regression, polynomial regression, pollutants, particulate matter, lasso regression, decision tree.