In recent decades decreasing quality of air and increased air pollution is one of the major problems faced by the world. In order to save the environment from further deterioration and curb the increasing air pollution it is important that informed policies are planned and executed. In order to do so, forecasting air pollutants is one of the most important tasks. In this study two models namely a deep LSTM neural network and the Facebook Prophet model are suggested for air quality criteria pollutants namely SO2, NO2 and PM10. These models are used to forecast the values of the criteria pollutants for Kanpur city, which is one of the most polluted cities in India. The results show that the LSTM model has higher performance accuracy and doesn’t overfit the forecasted trends whereas the prophet model due to its additive nature tends to overfit the trends when there is a clear upward or downward trend present in the time series. Both the models can be fine tweaked with expert advice and stand to perform better with a more consistent dataset.