An Improved Air Quality Index Machine Learning-Based Forecasting with Multivariate Data Imputation Approach

Accurate, timely air quality index (AQI) forecasting helps industries in selecting the most suitable pollution control measures and public reducing harmful exposure to pollution. This article proposes a comprehensive method forecast AQIs. Initially, work focused on predicting hourly ambient concentrations of PM2.5 PM10 using artificial neural networks. Once was developed, extended prediction other criteria pollutants, i.e., O3, SO2, NO2, CO, which fed into process estimating AQI. The AQI not only requires selection robust model, it also heavily relies sequence pre-processing steps select predictors handle different issues data, including gaps. presented dealt with this by imputing missing entries missForest, machine learning-based imputation technique employed random forest (RF) algorithm. Unlike usual practice RF at final stage, we utilized data feature selection, obtained promising results. effectiveness examined against linear for six pollutants proposed approach validated observations Al-Jahra, major city Kuwait. Results showed that models trained missForest-imputed could generalize accuracy 92.41% when tested new unseen is better than earlier findings.