Research On Temporal Pattern Mining Algorithm Based On Interval Events

The problem of mining frequent patterns from interval-based events is studied. It is assumed that the database consists of sequences of events, where an event occurs during a time-interval. The goal is to mine temporal arrangements of event intervals that appear frequently in the database. The motivation of this work is the observation that in practice most events are not instantaneous but occur over a period of time. Thus, there are many practical applications that require mining such temporal correlations between intervals including the network monitoring, data analysis of marketing records from supermarkets or medical dataset.We improved an existing algorithm called EMEMISP (Extending of MEMory Indexing for Sequential Pattern mining) for the discovery of the temporal patterns from interval data. The MEMISP algorithm is more efficient than both GSP and PrefixSpan algorithms in finding sequential patterns from transactional databases, which the EMEMISP algorithm choose to extend for mining frequent temporal patterns based on interval events. Compared to EMEMISP algorithm, two improved aspects were putted forward in our algorithm. First a pruning was applied during the mining process according to the Apriori principle, which can reduce the cost of computation. The other aspect is to reduce storage space by using pointers which point to relative frequent 2-patters instead of storage all the relations between each two events.Similar to EMEMISP algorithm, our algorithm also requires one database scan and does not require candidate generation or database projection, and after discovering all frequent temporal patterns, we show the method to generate temporal rules from the frequent patterns.Except the improvement of EMEMISP algorithm, we also researched some key programs that affect the process time of the temporal pattern mining algorithm.