Analyzing Hybrid Machine Learning Performance Through Air Quality Prediction

This paper focuses on machine learning approaches to analyze performance metrics based on air quality index prediction. We know that air pollution is one of the most serious problems faced by mankind to date and has severe consequences. The need to comprehend pollution patterns and accurately predict pollution levels will help communities take informed measures to mitigate its impact. This study begins with traditional machine learning approaches as the preliminary analysis and progresses toward more advanced hybrid models. This entire methodology aims to reduce or minimize the biases and overfitting of just one model by utilizing the properties of several models, thus improving AQI prediction accuracy. We compare the models on MSE, MAE and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R^{2}$</tex> metrics and propose two models as they display the best overall performance. Our proposed hybrid models Model1 and Model2 attain one of the best <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R^{2}$</tex> value of 0.851680 and 0.848379, and MSE values of 2200.807 and 2249.7899 respectively. Model1 is defined as RandomForestRegressor + ExtraTreesRegressor and Model2 as RandomForestRegressor + AdaBoostRegressor + XGBRegressor + ExtraTreesRegressor (Method of Stacking + Averaging). We do need to further evaluate these models with hyperparameter tuning to see how they change the results, however, we believe that these two hybrids, especially the latter, has a greater ability to generalize and obtain a near accurate prediction.