Particle Swarm Optimization based K-Prototype Clustering Algorithm

Clustering in data mining is a discovery process that groups a set of data so as to maximize the intracluster similarity and to minimize the inter-cluster similarity. The K-Means algorithm is best suited for clustering large numeric data sets when at possess only numeric values. The K-Modes extends to the K-Means when the domain is categorical. But in some applications, data objects are described by both numeric and categorical features. The K-Prototype algorithm is one of the most important algorithms for clustering this type of data. This algorithm produces locally optimal solution that dependent on the initial prototypes and order of object in the data. Particle Swarm Optimization is one of the simple optimization techniques, which can be effectively implemented to enhance the clustering results. But discrete or binary Particle Swarm Optimization mechanisms are useful for handle mixed data set. This leads to a better cost evaluation in the description space and subsequently enhanced processing of mixed data by the Particle Swarm Optimization. This paper proposes a new variant of binary Particle Swarm Optimization and K-Prototype algorithms to reach global optimal solution for clustering optimization problem. The proposed algorithm is implemented and evaluated on standard benchmark dataset taken from UCI machine learning repository. The comparative analysis proved that Particle Swarm based on K-Prototype algorithm provides better performance than the traditional K-modes and KPrototype algorithms.