Research Of Clustering Algorithm With Artificial Bee Colony

Data mining is the analysis methods of discovery of the potential value ofinformation in a database, it is a relatively young and interdisciplinary field of computerscience, also is the process finding a new model from the large amounts of data, it’sinvolved in artificial intelligence, machine learning, statisticaland database systems.The overall objective of the data mining process is extracted from a data structure thatcan be understood from a set of useful information, including the steps of the originaldata analysis and database and data management. Data mining does not require priorknowledge of data-related knowledge, through the analysis of the data, the object datadivided into different classes, a high similarity between the data which are in the samegroup, while the data which are in different groups should have a high degree ofdistinction from each other.The artificial bee colony is a swarm intelligence optimization algorithm based onthe definition of the intelligent behavior of honey bees. Its main feature is unnecessaryto know the specific information only need comparing the merits of the issues, byindividual local searching behavior of the worker bees, and ultimately find the globaloptimum in groups. Fast convergence, easy parallel implementation and wideapplication in many fields is also its advantages. Given these advantages, combinedwith foraging behavior of bees and the idea of clustering, proposing some views andimprovements.In artificial bee colony algorithm, the process of finding high-fitness food source isthe process of seeking the optimal solution. While in the artificial bee colony clusteringalgorithm, the process of finding the high adaptation of the food source is the process offinding the optimal cluster centers.Artificial bee colony clustering algorithm implemented in this article define thecluster centers as the food sources, two indicators for every cluster: the quality ofpolymerization, the quality of dispersion which according to is defined as the objectivefunction fi of each food source, and then calculated for each food source of fitness.In our experiment, we used the three standard data sets for testing, the resultsshowed that some indicators of the proposed method are better than the existedclustering algorithm based on artificial bee colony clustering and classical K-means,KHM algorithm.