Improving air quality prediction using hybrid BPSO with BWAO for feature selection and hyperparameters optimization.

Area of Science:

• Environmental Science
• Data Science
• Machine Learning

Background:

• Air pollution presents a substantial risk to public health and environmental sustainability.
• Accurate air quality prediction models are essential for effective environmental management and public health protection.

Purpose of the Study:

• To develop and optimize machine learning models for forecasting air quality using annual AQI data.
• To identify the most relevant features for AQI prediction through advanced feature selection techniques.
• To enhance model accuracy and computational efficiency via hyperparameter optimization.

Main Methods:

• Utilized annual Air Quality Index (AQI) dataset from the U.S. Environmental Protection Agency (EPA).
• Applied Binary Grey Wolf Optimizer (BGWO), Particle Swarm Optimization (BPSO), Whale Optimization Algorithm (BWAO), and a hybrid BPSO-BWAO for feature selection.
• Evaluated various machine learning models including Random Forest (RF), Gradient Boosting (GB), K-Nearest Neighbors (KNN), Multi-Layer Perceptron (MLP), Support Vector Machine (SVM), and Linear Regression (LR).
• Implemented hyperparameter tuning using a novel hybrid PSO-WAO approach.

Main Results:

• The hybrid BPSO-BWAO feature selection method demonstrated improved stability and feature set balance, selecting key AQI indicators.
• The Random Forest model, after feature selection, achieved an MSE of 53.93 and an R² of 0.9710, with reduced fitting time.
• Further optimization with the hybrid PSO-WAO enhanced the Random Forest model, resulting in an improved MSE of 51.82 and an R² of 0.9821.

Conclusions:

• Feature selection and hyperparameter optimization are critical for improving the accuracy and computational efficiency of air quality forecasting models.
• The optimized Random Forest model provides a robust framework for reliable air quality prediction.
• The study highlights the effectiveness of advanced optimization techniques in environmental data analysis.