摘要： 针对模糊C 均值算法需要不断迭代来计算样本数据的隶属度值以及聚类中心的特点,利用MapReduce 模型解决海量数据下的模糊C 均值问题,进而提出高效的模糊C 均值算法。在Map 阶段和Reduce 阶段分别完成隶属度和聚类中心的计算,每次迭代都需要启动一次完整的MapReduce 执行过程。通过多次迭代计算出隶属度值以及聚类中心,并更新聚类中心文件,供下一轮作业使用,重复执行这一过程直至得到最终聚类结果。实验结果表明,该算法能够有效减少MapReduce 计算过程中的迭代次数,从而提高整体执行效率。

关键词: 模糊C 均值算法, MapReduce 模型, 海量数据, 高效, 迭代

Abstract: Fuzzy C-means(FCM) algorithm requires constant iteration to calculate the characteristics of the membership value of the sample data and cluster center,using MapReduce model to solve the FCM under massive data. Map stage calculates membership degree,and Reduce stage completes computing cluster center. Each iteration needs to start a MapReduce implementation process. Through multiple iterations,it calculates the value of membership and cluster center, and updates cluster center file for the use of next round job. Repeat this process until get the final clustering results. Experimental results show that the algorithm can effectively reduce the number of iterations during the calculation and improve the overall efficiency of the implementation.

Key words: Fuzzy C-means(FCM) algorithm, MapReduce model, mass data, high efficiency, iteration

中图分类号: 

• TP391． 41