Multiobjective Interactive Fuzzy Clustering for Gene Expression Data

Clustering, an unsupervised method for classifying patterns, seeks to organize data points based on their similarities or differences. Common clustering techniques like K-means and Fuzzy C-means often face challenges with local optima, as they focus on optimizing a single cluster validity index. To address this, Genetic Algorithms (GAs) have been used, but their performance across different datasets remains inconsistent. This chapter introduces a new method called Interactive Multiobjective Clustering (IMOC), which adapts and optimizes objective functions during execution. IMOC involves a human decision-maker (DM) in the evaluation process, starting with an initial set of cluster validity indices. To manage DM fatigue, IMOC selectively presents solutions for evaluation, guided by the Non-dominated Sorting Genetic Algorithm II (NSGA-II) and Visual Analysis for Cluster Tendency Assessment (VAT) plots. Evaluations on real-life microarray gene expression datasets, including ”human fibroblasts serum” and ”yeast cell cycle” data, demonstrate IMOC’s superiority over traditional methods like K-means and FCM. Statistical tests confirm IMOC’s significant performance improvement, positioning it as a robust solution for enhancing clustering outcomes across diverse domains.