An Improved Initialization Method using Firefly Movement and Light Intensity for Better Clustering Performance

K-Means and EM algorithms are the most well-known clustering algorithms because they are simple, easy to understand and implement. However, both algorithms are sensitive to initial seeds which are randomly selected leading to slow convergence and less reliable clustering results. In this paper, an improved initialization method adopted the concept of light intensity and firefly movement to search for better initial seeds. Numerical experiments were conducted to evaluate the performance of the Enhanced K-Means and EM using faculty performance evaluation ratings as the dataset. The experiments showed that the implementation of the improved initialization method before the clustering process resulted in a higher intra-cluster and lower inter-cluster similarity. Also, there are fifty-nine percent (59%) and sixty-three percent (63%) decrease in the runtime execution while there are forty-four percent (44%) and twenty-seven percent (27%) fewer number of iterations recorded in the implementation of the enhanced KMeans and EM algorithms respectively.