Multiobjective Evolutionary Optimization for Prototype-Based Fuzzy Classifiers

Evolving intelligent systems (EISs), particularly, the zero-order ones have demonstrated strong performance on many real-world problems concerning data stream classification while offering high model transparency and interpretability thanks to their prototype-based nature. Zero-order EISs typically learn prototypes by clustering streaming online in a “one pass” manner for greater computation efficiency. However, such identified often lack optimality, resulting less precise boundaries, thereby hindering potential of systems. To address this issue, commonly adopted strategy is minimize training error models historical or alternatively iteratively intracluster variance clusters obtained via partitioning. This recognizes fact that ultimate driven positions space. Yet, simply minimizing may potentially lead overfitting does not necessarily ensure optimized attain improved outcomes. achieve better avoiding EISs, article presents novel multiobjective optimization approach, enabling obtain optimal involving these two disparate but complementary strategies simultaneously. Five decision-making schemes are introduced selecting suitable solution deploy from final nondominated set models. Systematic experimental studies carried out demonstrate effectiveness proposed approach improving EISs.