Thesis (BS Computer Science) -- University of the Philippines Mindanao, 2011

Data clustering is the unsupervised classification of unlabeled data objects into groups called clusters. It is one of the most primitive activities of human beings, and has been extensively used for understanding and utility. One type of clustering is K-means clustering, where data objects are partitioned int multiple clusters. This paper proposed a new approach in solving the K-means clustering problem using a novel hybrid of Continuous Tabu Search (CTS) and a modified Firefly Algorithm (FA). The new algorithm, called Continuous Tabu-Firefly Algortihm (CTFA), used the CTS as a local search method embedded in the move operator of the modified FA. CTFA was tested against the pure Firefly Algorithm and the Hybrid K-means and Particle Swarm Optimization. The performance of each algorithm was benchmarked using the Iris and Wine data sets. The results of the study show that CTFA was able to surpass the clustering efficiency of both algorithms in terms of solution quality. With regards to solution time, CTFA took longer to generate the solution. However, CTFA still has shorter solution time compared to other brute force methods.