模糊时态序列演化模式挖掘

Fuzzy logic for mining of temporal sequence evolution patterns

目前时态序列挖掘方法大多都是以一种自然的方式对序列分割、离散处理等,从而使离散化结果很大程度依赖于外部的人为分割变量。为了使离散化结果更强地依赖于原始数据,应用模糊聚类方法,将连续时态演化序列转变为模糊时态演化序列,应用模糊时态演化片段支持度评定频繁模糊时态演化模式,用隶属度计算关联规则的支持度和可信度,使这两个重要指标计算更为精确。给出了频繁模糊模式集的生成算法和复杂度。实际算例显示了方法的有效性。

Most existing temporal sequence mining methods depend on partitioning and discretization in a natural way, which brings on that the symbols of the alphabet are usually chosen externally and imposed by the users.In order to reduce the randomicity, the original temporal sequence is transformed into fuzzy form by fuzzy clustering, and then frequent fuzzy temporal evolution patterns are assessed with support and confidence measure.Rule＇s support and confidence are calculated from membership and each sample does not arbitrarily support a single symbol so as to make the two important measures more exact and actual.An apriori algorithm and its complexity for discovering frequent fuzzy itemsets are present.The practical cal- culation shows that the mining of temporal sequence evolution patterns on commodity futures data is meaningful and resultful.