An Online Predictor Model as Adaptive Linear model and Evolving Takagi-Sugeno Model

摘要

In this paper, we suggest an online predictor model to follow the variations of a process with significant uncertainties. This model is defined as an adaptive linear model initially; however, it starts to evolve into a TS model and continues the process while the modeling error is higher than an adaptive threshold. The model habitually returns to an adaptive linear model when the modeling error becomes lower than the adaptive threshold. Evolving from a linear model to a nonlinear one and then returning to a linear model is performed through specially designed split and merge procedures. Finally, a real-world case study is conducted: prediction of monthly number of sunspots demonstrates the capabilities of the proposed approach.