Air pollution is one of the most critical risk factors for human health. Exposure to high concentrations of PM2. 5, NO, NO2, O3, and PM10 is a significant cause of death from cardiorespiratory conditions. Preventive measures can mitigate the adverse effects of air pollution if predicted in advance. The existing sequential methods are not able to handle the huge volume of data generated by the sensors to forecast the results in a timely manner. Therefore, this paper introduces a novel method (GOA-SSVR) that leverages Apache Spark’s strength, grasshopper optimizer, and support vector regression to forecast NO, NO2, O3, PM10, and PM2. 5 concentrations 24 hours in advance. The proposed method employs Apache Spark to create a parallel architecture, whereas grasshopper optimizer is used to optimize the hyperparameters of SVM. To test the performance, root mean squared error (RMSE), mean absolute error (MAE), and \ (R²\) values of the proposed and other considered methods have been compared. The proposed model achieves notable quantitative enhancements, with a 25% reduction in mean RMSE and an 18% decrease in mean MAE for combined meteorological variable sets (TMV+TPC+HPC), alongside a 10% improvement in mean \ (R²\) compared to traditional baseline methods. To ensure the reliability of the results, statistical validation methods, including the Friedman test followed by the Least Significant Difference (LSD) test, were employed. Furthermore, the integration of Apache Spark significantly improves computational efficiency, reducing execution time by 28% on large datasets, enabling scalable and faster processing. These findings demonstrate GOA-SSVR’s enhanced predictive accuracy and computational efficiency, affirming its practical utility in large-scale air quality forecasting.
Tripathi et al. (Sat,) studied this question.