To Identify Dynamic Behaviour of Frequent Patterns by Exploiting Timestamps

Mining frequent item-sets or patterns from an online transactional database is one of the fundamental and essential operations in many data mining applications. Apriori and FP-growth are some of the examples for the existing frequent pattern mining algorithms. Only the frequent items sets and their counts are found out by these algorithms. They do not consider anything about the time stamps associated with the transaction. Each transaction database usually consists of time stamp of each transaction. This time stamp is a sequence of characters denoting the date and time at which a certain event occurred. This paper extends the existing frequent pattern mining algorithms to take into account time stamp of each transaction. And also discovers a new type of patterns whose frequency dramatically changes over time which is defined as transitional patterns. The frequency of the transitional patterns may increase or decrease at some time points in a transaction database. These patterns capture the dynamic behavior of frequent patterns in a transaction database. This paper also studies a new concept called significant milestones, which are time points at which the frequency of the pattern changes most significantly. This paper objective is to find out such transitional patterns and their significant milestones that considering the timestamp of each transaction, a modified transitional pattern mining algorithm is presented.