CAJ | 학술논문

针对聚类算法并行化的需求，该文对基于Hadoop平台Kmeans算法进行了改进，选用Canopy算法对数据进行预处理，并在具有一定数据结构的电影数据集上进行了单机对比实验，集群加速比实验和集群扩展率实验，分别体现改进后算法实现的高效性、良好的加速比和可扩展性，从而可以有效地运用在实际海量数据挖掘中．
침대취류산법병행화적수구，해문대기우Hadoop평태Kmeans산법진행료개진，선용Canopy산법대수거진행예처리，병재구유일정수거결구적전영수거집상진행료단궤대비실험，집군가속비실험화집군확전솔실험，분별체현개진후산법실현적고효성、량호적가속비화가확전성，종이가이유효지운용재실제해량수거알굴중．
According to parallelism demand of the clustering algorithm, This paper improved the implemention of the kmeans algorithm based on the Hadoop platform. We do the preprocess on the dataset using the canopy algorithm, and conduct the single contrast experiment, cluster speed up experiment and cluster expansion rate experiment, showing the high effiency, better speed up and scalability, thus the implemention can be used in the pratical mass data mining effectively.