基于滑动窗口的支持泛在应用的流聚类挖掘算法

Clustering Algorithm for Ubiquitous Data Stream Mining Based on Slip Window

近年来,泛在数据流挖掘逐渐成为数据挖掘发展的新热点,它具有在有限的资源上去挖掘无限的数据流,并可随时随地返回挖掘结果的特点,对此,本文提出一种基于滑动窗口的流聚类算法;该方法将一个滑动窗口分成n个大小相等的窗口单元,基于窗口单元进行增量式的知识相关性的挖掘,提高了流挖掘的效率;当窗口滑动时,通过衰变函数衰减当前滑动窗口内的第一个窗口单元的挖掘结果,并在当前滑动窗口挖掘结果中将其剔除,实现下一滑动窗口的增量式挖掘.

Recently, ubiquitous Data Stream Mining has become a new focus of data mining gradually. For its having finite ubiquitous resource mine infinite data stream and returning outcome anytime and anywhere,this paper suggest stream clustering algorithm based on slip window. It divides a window into n average units, on which the increasable and knowledge-correlated mining is executed based, so as to heighten its efficiency. As window is to slip,algorithm weaken the first unit＇s outcome of current window with weakened function and eliminate its affect for current outcome so as to make true the next increasable mining in next window.