Meta-cognitive Interval Type-2 neuro-fuzzy inference system for wind prediction

In this paper, we propose an Interval Type-2 neuro-fuzzy inference system and its meta-cognitive projection based learning algorithm (PBL-McIT2FIS) for wind speed prediction. Interval Type-2 fuzzy sets are employed in the antecedent of fuzzy rules and the consequent realizes Takagi-Sugeno-Kang (TSK) inference mechanism. Initially the rule base in PBL-McIT2FIS is empty, the learning algorithm employs prediction error and novelty of sample as a measure to add rules to network. As each sample is presented to network, the meta-cognitive component decides on whether to delete the sample without learning, learn the sample by adding a new rule, update the existing rules or reserve the sample for future use. Whenever a new rule is added or parameters of existing rules are updated, a projection based learning algorithm is employed to compute the optimal weights of the network. Performance of PBL-McIT2FIS is evaluated on a real world wind prediction problem and compared with support vector regression and OS-fuzzy-ELM. The results indicate better performance of PBL-McIT2FIS.