Mamdani-Type Fuzzy Modelling via Hierarchical Clustering and Multi-Objective Particle Swarm Optimisation (FM-HCPSO)

In this paper, a systematic data-driven fuzzy modelling approach is proposed, which integrates transparent fuzzy systems (linguistic fuzzy systems) with an effective evolutionary computing based algorithm - the new structure Particle Swarm Optimisation (nPSO). In this modelling mechanism, a new data clustering technique via an improved hierarchical clustering algorithm is designed for the initial fuzzy model generation. Multi-objective optimisation techniques are then employed for the improvement of the generated fuzzy model, which takes into account both the accuracy and the interpretability performances of the fuzzy system. This proposed modelling approach is tested on two benchmark problems and a high-dimensional modelling problem using real industrial data. This latter concerns the prediction of the mechanical properties of alloy steels. Experimental results show that the proposed approach is very effective in eliciting accurate as well as interpretable Mamdani-type fuzzy models.