First order of autoregressive air pollution forecasting with symmetry triangular fuzzy number based on percentage error

Abstract

Autoregressive (AR) models is known best to predict multiple sets of stationary data. Previous AR model uses single point data, though uncertainties does exist in data due to various factor. When the data contain uncertainty, traditional procedure which is developed to handle the single point (crisp) data is insufficient to deal with the uncertain data. Moreover, unresolved uncertainty in data may increase error in prediction model. That is, data collected that contains uncertainty should be adequately treated before being used for analysis. Hence, this study proposes an first order of autoregressive (AR(1)) model building based on symmetry triangular fuzzy number. The triangles are established from percentage error method during data preparation of AR(1) modelling to address the uncertainty issue. In this study, AR(1) model with fuzzy data is built to forecast air pollution. The result of this study demonstrates that the proposed method of building fuzzy triangles for AR(1) model obtain smaller error in prediction. The improvement on the existing data preparation process sought from this study is expected to give benefit in achieving better forecasting accuracy and dealing with uncertainty in the analysis.