Statistical methods are appropriate for the prediction around a median, but not around a quarter or three quarters. On the prediction of environmental phenomena, it is important to predict correctly at high concentrations and to reasonably explain the predictive formula.
In this paper, we propose a method to predict phenomena composed of many complicated factors by modeling the process of human thinking and judgement. Input-output relations of the system are described in the form of if-then rules. Then using fuzzy reasoning, the behaviors of the system will be predicted. To improve the model, we adjust the parameters of the membership function, by the use of the non-linear optimization technique.
We apply fuzzy modeling to the prediction of oxidant concentrations at Osaka area in Japan and show that it is appropriate for the prediction of phenomena with limited input-output data.
抄録全体を表示