Type-2 fuzzy rule base system with parameter optimization for forecasting of tardiness

This paper addresses an interval type-2 fuzzy hybrid rule-based system in order to predict the amount of tardiness where tardiness variables are represented by interval type-2 membership functions. For this purpose, interval type-2 fuzzy disjunctive normal forms and fuzzy conjunctive normal forms are utilized in the inference engine. The main contribution of this paper is to present the interval type-2 fuzzy hybrid rule-based system, which is the combination of Mamdani and Sugeno methods. In order to forecast the future amount of tardiness for continuous casting operation in a steel company in Canada, an autoregressive moving average model is used in the consequents of the rules. Parameters of the system are optimized by applying Adaptive-Network-Based Fuzzy Inference System (ANFIS). This method is compared with interval type-2 fuzzy Takagi-Sugeno-Kang method in MATLAB, multiple-regression, and two other Type-1 fuzzy methods in literature. The results of computing the mean square error of these methods show that our proposed method has less error and high accuracy in comparison with other methods.