Air Pollution Forecasting with Random Forest Time Series Analysis

Abstract

Air pollution is increasing day by day in the metropolitan area. In this paper, two important air pollution indicators, particulate matter 10 (PM10) and carbon monoxide (CO), are forecasted with random forest time series analysis and artificial neural networks method using meteorological data such as air temperature, humidity, wind speed and air pollutant data as input for Istanbul - Sisli and Kocaeli - Dilovasi areas. Air pollution forecasts are made for the following day with data gathered from Republic Of Turkey Ministry Of Environment's and Turkish State Meteorological Service's after pre-processing. When forecasting results are examined, it is clearly seen that random forest method produces very accurate results and performs better than artificial neural networks.