PM10 levels are a key problem for Mandideep’s air quality. We show a collection of results acquired by applying various neural network algorithms to the PM10 prediction problem. The 1-day ahead forecast has an acceptable level of accuracy, which can be enhanced further if a proper deseasonalization strategy is used, transferring some prior knowledge during the data preprocessing step. The prediction shows the best output considering MLP-3 model. A drop in accuracy was observed while using the statistical one. The study shows to have a considerable correlation with a huge set of meteorological parameters. It is evident from the study that the input variables set comprises the concentration of PM10, NOx, SO2, and a wide set of meteorological variables including ambient temperature, humidity, wind velocity, and solar radiation can give an accurate prediction of concentration of PM10 in a day advance.