Air pollution level prediction system

Abstract

Nowadays, the levels of air pollutants in the environment are increasing manifold. This has led to deterioration of human lifestyle. Various methods like ‘Climatology’ (based on the assumption that the past is a good predictor of the future) have been used for air quality forecasting. These approaches are usually used to predict exceeding limits from specific thresholds, not ambient concentrations. As a result, a lot of improvement is still required in this field for prediction analysis. With incomplete data parameters and their significance (priority), most of the methods fail to predict the pollution levels significantly. The advantage of artificial neural networks includes the problem-solving efficiency in the cases of unavailability of complete information, with no information about the analytical relationship among the input and processed output data. The aim is to develop an artificial neural network for air quality prediction that can perform with constrained dataset with highly robust feature in order to handle the data including noise and errors. Dataset used deals with pollution in the U.S. involving four major pollutants (Nitrogen Dioxide, Sulphur Dioxide, Carbon Monoxide and Ozone) on daily basis for the time period of year 2008 to 2017. We use prediction models like ARIMA etc. to validate our predicted AQI. This AQI analysis helps in telling the status of present air pollution and forecasted pollution levels in coming time. So, it plays a vital role for decision maker and for individual also to know about air pollution quality.