Variable strategy ensemble artificial bee colony algorithm for automatic data clustering
Vaishali Pavalbhai Patel, Ashish Tiwari, Amit S. Patel · 2016
Data analysis is a challenging task in this information world considering this complexity clustering play a significant role in data mining. Many bio-inspired algorithms are applied to clustering application. But during execution of these algorithms it's required number of cluster at initial stage, and in real life application it is difficult to identify correct number of cluster from input data. This paper points out this deficiency by introducing new automatic clustering algorithm with ABC (artificial bee colony optimization). And it is accomplished by Euclidian distance and statically properties measure in terms of selection threshold and threshold cut off. In addition to this weighted Euclidian distance is used to assign data to different cluster centre. The ABC algorithm has many real applications in solving an optimization problem. But search equation performed by employed and onlookers bees is heavily depend on random search which have effect in terms of sufficient at exploration but insufficient at exploitation. To reduce this limitation, in this work, new food search strategy based on exploitation mechanism of PSO is proposed for employed bee. In the proposed search strategy new position of employed bee depends on the global best from the population as well as the local best of the current solutions. The developed method is able to find any complex cluster irrespective of their data distribution, density, shape and type. It is compared with existing well-proven automatic clustering techniques. The performance results prove that the proposed technique give a more accurate number of cluster and converge faster than other.