针对数据对象在数据流中的频繁度变化趋势的预测问题,提出基于最大最小频率时间窗模型的最大最小频繁趋势预测算法(MM-FTP).设计一种新的最大最小频繁模式树结构(MMFP-Tree),存储数据流概要信息;提出一种新的数据对象频繁度变化趋势衡量指标——频繁度变化率(FCR),定量地对数据对象的频繁度变化趋势进行描述.该算法同样能够对数据流分类置信度变化趋势及传统的指数变化趋势进行有效预测.结果表明,在真实的网络点击数据流上,该算法能够快速准确地预测数据对象的频繁度变化趋势.
For the frequency tendency prediction problem of itemsets over streams, a novel max-rain frequency tendency prediction (MM-FTP) algorithm is proposed based on the Max-Min Frequency Window model. A max-rain frequency pattern Tree (MMFP-Tree) structure is established to store the summary information of streams a new measure frequency changing rate (FCR) is presented to describe the frequency tendency of itemsets quantitatively. The MM-FTP algorithm is useful in the index tendency prediction and the confidence prediction of classification. Based on the result of the case study on web log data stream, the MM-FTP algorithm could be used to predict the frequency tendency efficiently and effectively.