Clustering Gene Expression Data Based on Harmony Search and K-harmonic Means

Clustering is one of the most commonly data explorer techniques in Data Mining. K-harmonic means clustering (KHM) is an extension of K-means (KM) and solves the problem of KM initialization using a built-in boosting function. However, it is also suffering from running into local optima. As a stochastic global optimization technique, harmony search (HS) can solve this problem. HS-based KHM, HSKHM not only helps KHM clustering escaping from local optima but also overcomes the shortcoming of slow convergence speed of HS. In this paper, we proposed a hybrid data-clustering algorithm, HSKHM. The experimental results on four real gene expression datasets indicate that HSKHM is superior KHM and HS in most cases. The HSKHM algorithm not only improves the convergence speed of HS but also helps KHM escaping from local optima.