@article{MAKHILLAJIT201615226505,
    title = {Hybridization of K-Means and Harmony Search Based on Optimized Kernel
Matrix and Unsupervised Constraints},
    journal = {Asian Journal of Information Technology},
    volume = {15},
    number = {22},
    pages = {4512-4521},
    year = {2016},
    issn = {1682-3915},
    doi = {ajit.2016.4512.4521},
    url = {https://makhillpublications.co/view-article.php?issn=1682-3915&doi=ajit.2016.4512.4521},
    author = {S. and},
    keywords = {Document clustering,K-means algorithm,harmony search,kernel function,particle swarm optimization},
    abstract = {Clustering is one of the effective techniques that separate the data into meaningful groups. Feature
selection is an important concept to enhance efficiency in clustering process. Existing work presented a method
called hybridization of K-means algorithm and Harmony Search Method (HSM) for clustering the documents.
In this method, concept factorization is used to extract the meanings to cluster the documents. But it needs to
improve clustering accuracy in the document clustering process. In this manuscript, Kernel and Weighted
feature based Clustering (KWC) method is presented to cluster the documents. Spherical kernel is utilized as
the higher order kernel that is higher rate of computation. Furthermore, the weight of each concept is calculated
and select as the weighted features. The problem in this method is poor generalization performance so it needs
to select optimal kernel matrix. So, Particle Swarm Optimization (PSO) based Optimal Kernel Matrix Selection
(PSO-OKMS) is presented to select the optimal value of kernel matrix. In this method, kernel set is to chosen
accurately to improve clustering performance but the accuracy is less. Furthermore Unsupervised Constrained
based Hybrid Clustering (UC-HC) to improve the clustering performance. In this method, data are extracted by
identifying an assignment that rises similarity score between strings and informs to the constraints.
Experimental result compares methods such as KWC, PSO-OKMS and UC-HC to measure the clustering
accuracy. The proposed UC-HC method shows high accuracy when compared to KWC and PSO-OKMS
methods.}
    }