Discovering Transitional Patterns and Their Significant Milestones in Transaction Databases

摘要: A transaction database usually consists of a set time-stamped transactions. Mining frequent patterns in databases has been studied extensively data mining research. However, most the existing pattern algorithms (such as Apriori and FP-growth) do not consider time stamps associated with In this paper, we extend framework to take into account stamp each discover whose frequency dramatically changes over time. We define new type patterns, called transitional capture dynamic behavior database. Transitional include both positive negative patterns. Their frequencies increase/decrease at some points introduce concept significant milestones for pattern, which are significantly. Moreover, develop an algorithm mine from along their milestones. Our experimental studies on real-world illustrate that is highly promising practical useful approach discovering novel interesting knowledge large databases.