Air quality is currently seen as a global health issue as a result of human activities and the accelerating industrialization over the past several years. Therefore, it is essential to accurately estimate air quality. In estimating the air quality of a certain area, machine learning has consistently demonstrated promising results. To forecast the numerous air particles that are PM2.5, PM10, NO2, NH3, SO2, CO, and ozone O3 concentration in different cities of Pakistan, the authors presented a variety of machine learning models, including linear regression (LR), Decision Tree (DT) using regressor and classifier, Random forest (RF) using regressor and classifier, K nearest neighbour (KNN), and the deep learning models LSTM and BILSTM. Additionally, a comparison of the various models is made to determine which is most important. The Random forest model and KNN are found to be the most dependable models for the findings with respect to mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R2). This study found that the suggested models anticipate the PM2.5 pollutants more accurately and with a lower error rate than the existing models after comparing a number of current models and new models.