Optimized Clustering Algorithm with Swarm Intelligence for Large Dataset

Abstract

The compactness and content validation of the clustering algorithm is a significant challenge. The partition-based clustering algorithms achieve the paramount of compactness and validation. The swarm-based clustering algorithm has great protentional over the large dataset. This paper presents the modification of K-means algorithms with particle swarm optimization and ant colony optimization. Both swarm algorithms enhance the performance of the K-means algorithm. The derive objective function convert into the fitness function of the swarm algorithm. The modified algorithm called KACO and KPSO implement in MATLAB software. To evaluate the performance of the modified algorithm, applied one real dataset and two synthetic datasets. The results of the modified algorithm are better than the FCM and K-means algorithm. The validation of cluster centers distance measure by different distance formulae such as Euclidean distance, cosine distance, and others. The applied distance formula influences the results of the clustering algorithm. (The optimal solution set of centers formed an optimal cluster.) For empirical evaluation, measure three parameters like the number of iterations, standard deviation, error rate. The minimization factor of validation parameters indicates the compactness and validation of the clustering algorithm.