New forecasting model using type-2 fuzzy multivariate time series

• Published in: 2011 National Postgraduate Conference pp. 1 - 4
• Main Authors: Hashim, M. A., Jaafar, J., Taib, S. M.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01-09-2011
• Subjects: Autoregressive processes; Forecasting; Fuzzy logic; fuzzy time series; fuzzy type-1; fuzzy type-2; Hidden Markov models; multivariate fuzzy time series; Predictive models; Time series analysis; Uncertainty
• ISBN: 1457718820; 9781457718823
• DOI: 10.1109/NatPC.2011.6136296

These days, fuzzy time series model has widely been applied in many applications to forecast about the future. In current fuzzy time series model only provides a single-point forecasted value and using less factors in prediction. Regarding the problems, many researches on the model have been done from time to time in order to find better accuracy forecasting models. This research aims to meet three objectives which are to understand and analyze current fuzzy time series development, to extend type-1 fuzzy to type-2 fuzzy and lastly to validate the accuracy of the new forecasting model. Fuzzy logic has solved the uncertainties problem that been faced by conventional time series models. However, the limitation of type-1 fuzzy in handling uncertainties has extended to type-2 fuzzy. The type-2 fuzzy has more degree of memberships that helps to handle more complex uncertainties values and provide more observations on data. Further research on fuzzy time series has brought towards the idea of multivariate fuzzy time series by more variables been used. By combining the type-2 fuzzy with multivariate time series will result more observations, more variables and also obtained more accuracy forecasting models.