PM2.5 is a hazardous form of air pollutant because of its diminutive size. Many Indian cities have experienced extremely high levels of PM2.5 in recent years. This study aims to examine the accuracy of various algorithms for forecasting PM2.5 category for the next day. Authorities and individuals alike might benefit from air quality forecasts that allow them to prepare for days with high pollution levels. Seven machine learning and one deep learning classification algorithms were employed to forecast the PM2.5 category under Indian NAAQS for three diverse regions in Gujarat, India. According to the results, the support vector machine (SVM) outperformed the deep learning algorithm and other machine learning algorithms with an average accuracy of 71.33%. Furthermore, the result suggests that more accurate outcomes can be obtained by reducing the number of categories used for categorization, such as two-category classification, good or bad.