Kernel Evolving Participatory Fuzzy Modeling for Time Series Forecasting: New Perspectives Based on Similarity Measures

The evolving Fuzzy Systems (eFSs) has demonstrated a powerful class of model for time series forecasting, due to their autonomy to handle with the data, and with highly complex problems in real-world applications. Unfortunately, real-world data contains sources of perturbations and variations that can be correlated to external variables, known as randomness. An alternative cause of randomness is chaos, forming the chaotic time series. This paper suggests the ePL-KRLS-FSM, a new class of evolving fuzzy modeling approach that combines participatory learning (PL), a kernel recursive least squares method (KRLS), and data transformation into fuzzy sets (FSs). This transformation allows better handling of inaccuracies with the data, proposing a model that can predict chaotic data with more exactness. The model is evaluated using the Mackey-Glass time series benchmark with different levels of chaos. Furthermore, the performance of the ePL-KRLS-FSM is compared with some related state-of-the-art rule-based eFSs and traditional forecasting models. The computational results show that the proposed model is competitive and performs more consistently than the compared models.