IoT Enabled Machine Learning for Vehicular Air Pollution Monitoring

Air pollution has become a real-life problem. One of the major pollutant emission is due to the increase in a number of vehicles. Key pollutants generated from the vehicles are Carbon Monoxide (CO), Oxides of Nitrogen (NOx) and Particulate Matter (PM). Conventionally to measure pollution from the road, simulation models were created or various expensive instruments were involved to measure pollution levels which are not likely to be used in real time. Here we have proposed an IOT-based model in which sensors were used to measure CO, PM pollution level and environmental condition like temperature and humidity. The main objective of this approach is to suggest an alternative route to the user based on pollution status and distance of each route which leads to a pollution-free route. The web-based application developed has a Google map API where the pollution status and alternative routes were suggested. With the collected time series samples, the prediction analysis was done for PM with neural network Multi-Layer perceptron and support vector machine regression (SVMR) learning algorithm. With the inferences from the prediction, it is proved that the neural network reduces Mean Absolute Error (MAE) by 27.27 and produces better accuracy when compared with SVMR.