The Maintenance of Frequent Episodes

Data mining technology is of great help in data analysis. Mining frequent episode is one of the important task in this field, which allows users to predict future events based on the current events. The traditional approaches for mining frequent episodes use hierarchical concept, that is, generate candidate episode first, and then scan the sequence data to determine whether they are frequent episode, is very time consuming to repeatedly scan the sequence data and search for candidate episodes. This paper proposes a method for mining episode in a data stream. Our method just scans new added data to update existing frequent episodes without scanning original data and searching for candidate episodes, which is more efficient than the other methods.