SVR machine learning and SARIMA-based air quality index classification and forecasting system

• Published in: Discover applied sciences Vol. 7; no. 9; p. 994
• Main Authors: Bahadure, Nilesh Bhaskarrao, Gonge, Sudhanshu, Parashar, Deepak, Shah, Bhoomi, Patil, Prasenjeet Damodar, Renugadevi, M., Raju, Nagrajan
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 26.08.2025; Springer Nature B.V
• Subjects: Accuracy; Air monitoring; Air pollution; Air quality; Algorithms; Applied and Technical Physics; Artificial intelligence; Autoregressive moving-average models; Chemistry/Food Science; Classification; Climate change; Clustering; Confidence intervals; Decision making; Earth Sciences; Ecosystems; Engineering; Environment; Environmental monitoring; International organizations; Learning algorithms; Machine learning; Materials Science; Nitrogen dioxide; Outdoor air quality; Perceptions; Performance enhancement; Pollutants; Pollution control; Pollution levels; Predictions; Public awareness; Quality management; Recall; Statistical analysis; Support vector machines; Sustainable development; Trends; VOCs; Volatile organic compounds; Seasonal autoregressive integrated moving average (SARIMA); Support vector regression; Time-series; Air quality index (AQI); Machine learning
• ISSN: 3004-9261; 2523-3963; 3004-9261; 2523-3971
• DOI: 10.1007/s42452-025-07327-0

The prediction and classification of the air quality index (AQI) are a primary concern. Still, due to the involvement of many parameters affecting AQI pre- dictions, it becomes time-consuming and monotonous work. The prediction and classification of the AQI with utmost accuracy is a pivotal tool for evaluating and monitoring the level of air pollution in a given area, facilitating public awareness and policy making in safeguarding human health and the surroundings. In this study, to improve the performance of air quality index classification and prediction, machine learning techniques based on support vector regression, and for prediction, the classic time-series analysis based on seasonal autoregressive integrated moving average (SARIMA) was employed. The forecasted AQI value for the next 25 days lies in a 97.3% confidence interval zone. The experimental results of the proposed method are evaluated and validated for performance and quality analysis on AQI data based on accuracy, precision, recall, and F1 score. The experimental results achieved 97% accuracy, compared to 94.1% with the Auto-regressive integrated moving average (ARIMA)-based technique, 91% precision, 94% recall, and 92% F1 score, demonstrating the effectiveness of the proposed technique for classifying and predicting the air quality index. Article Highlights Machine learning improves AQI prediction, making it faster and more accurate for public awareness and policy decisions. The model forecasts AQI with 97.3% confidence for 25 days, aiding proactive air quality management. Achieved 97% accuracy, proving the method’s reliability for classifying and predicting air pollution levels.