Journal of Applied Sciences1812-56541812-5662orgz10.3923/jas.2014.426.435AbuhamdahAnmarKittanehRawnaqAlkhatibJawadNazriMohd Zakree Ahmad52014145Clustering problem is an important area of data mining. The
goal of the clustering is to partitioning N objects into K clusters with similar
characteristics for one cluster and dissimilar to other clusters by using any
algorithm to generate the initial clusters partition. This initial clusters
quality will be measured based on a certain distance (or object) function. The
initial partitioning can be improve by iteratively explores its neighbor solutions
looking for a better one by any algorithm. The neighbor’s
solutions are achieved by reallocating some patters (objects) by using any neighborhoods
structures. However, the way of employing the neighborhoods structures will
affects the solution quality. Therefore, in this work, a swap mechanism is proposed
to increase the diversification for the neighborhood structure. The proposed
mechanism tested on two algorithms previously implemented over six medical clustering
problem (that are available in UCI Machine Learning Repository) by using two
way of the distance calculation. This work, also present a third distance calculation
to minimize the distances between the clusters. Experimental results show that,
using the swap mechanism and the third distance function on the algorithms is
able to produce significantly better quality solutions than not using it on
all datasets.]]>Abuhamdah, A.,2012Abuhamdah, A., B.M. El-Zaghmouri, A. Quteishat and R. Kittaneh,2012Berry, M.J. and G.S. Linoff,1997Brucker, P.,1978Dasgupta, S. and Y. Freund,2009Davidson, I. and A. Satyanarayana,2003Halkidi, M., Y. Batistakis and M. Vazirgiannis,2001Holland, J.,1975Hong, S.,2006Jain, A.K., M.N. Murty and P.J. Flynn,1999Mahajan, M., P. Nimbhorkar and K. Varadarajan,2009Saha, I., D. Pewczynski, U. Maulik and S. Bandyopadhyay,2010Wang, X.,2006Liu, Y., Z. Yi, H. Wu, M. Ye and K. Chen,2008Zhang, C., D. Ouyang and J. Ning,2010