Recommender systems are considered to be the most effective tools to provide suitable recommendations to the users that perform online transactions. The number of customers, their services and the need for online information has been increasing steadily. Clustering has been defined as a task that divides datasets so that the elements in a subset will be similar to each other and will differ with the elements of other subsets. This phenomenon may be further considered as a problem for looking up an ideal configuration among clusters of different configurations. The most predominant technique for approximation is K-means coming to the problem of clustering. Even though it is very simple and efficient, it is susceptible to its initial selection of cluster centers and may get stuck up within the local optima. There are potential meta-heuristic methods to deal with these issues. Direct application of Meta heuristics for clustering is practically valid only on smaller datasets. This work proposes methods to find the initial solution to K-means algorithm by using the Firefly Algorithm (FA) along with a Simulated Annealing (SA) algorithm that generates better solutions for search found near its global optima. The results of the experiment prove that the suggested FA-SA with k-means technique proved to achieve a performance that was better compared to SA alone.
Volume 11 | Issue 6