Modeling and Short-Term Prediction of Air Pollution by a Revised GMDH
Hiroyuki Tamura, Tadashi Kondo · Transactions of the Society of Instrument and Control Engineers · 1979
This paper deals with nonlinear statistical modeling of air pollution concentration for short-term prediction. The method used for modeling is a revised version of the basic GMDH (Group Method of Data Handling) algorithm. By using the time series data of the SO2 concentration, the wind velocity and the wind direction in Tokushima, Japan, we intend to find a suitable model for predicting SO2 concentration a few hours in advance. Firstly, the suitable data length for modeling air pollution in Tokushima is investigated. Secondly, three different prediction models obtained by the revised GMDH are compared to find a suitable structure and suitable input variables in the model. The predicted results obtained by the revised GMDH model are compared with the results obtained by a linear regression model, a linear autoregressive model and a basic GMDH model. It is shown that the revised GMDH model developed in this paper gives a better result and compares favorably with linear models and the basic GMDH model. It is also shown that the revised GMDH model obtained is much simpler than the basic GMDH model.