ABC_Kmeans聚类算法的MapReduce并行化研究

Map Reduce Parallel Study of ABC_Kmeans Clustering Algorithm

摘要:

随着数据的海量增长,数据聚类算法的研究面临着海量数据挖掘和处理的挑战；针对K-means聚类算法对初始聚类中心的依赖性太强、全局搜索能力也差等缺点,将一种改进的人工蜂群算法与K-means算法相结合,提出了ABC_Kmeans聚类算法,以提高聚类的性能；为了提高聚类算法处理海量数据的能力,采用MapReduce模型对ABC_Kmeans进行并行化处理,分别设计了Map、Combine和Reduce函数；通过在多个海量数据集上进行实验,表明ABC_Kmeans算法的并行化设计具有良好的加速比和扩展性,适用于当今海量数据的挖掘和处理。

Abstract:

With the massive growth of data, data clustering algorithm research is facing the challenge of mass data mining and processing.For K - means clustering algorithm to the dependence of the initial clustering center is too strong, and poor global search ability shortcomings, will be an improved artificial colony algorithm combined with K - means algorithm, ABC_Kmeans clustering algorithm is proposed, in order to improve the performance of clustering.In order to improve the clustering algorithm’s ability to deal with huge amounts of data, uses the MapReduce model for parallel processing to ABC_Kmeans, design the Map, Combine and Reduce function respectively,Through the experiments on several huge amounts of data collection,show ABC_Kmeans parallel design of algorithm has good speedup and scalability, applicable to today's huge amounts of data mining and processing.