An Efficient Approach for Maintaining and Mining Frequent Episodes

Data mining technology plays a crucial role in the area of data analysis. Mining frequent episodes stands out as a pivotal task within this domain, enabling users to forecast future events based on present occurrences. Conventional methods for discovering frequent episodes typically follow a hierarchical approach, involving the generation of candidate episodes and subsequent scanning of sequence data to count their frequency, which can be quite time-consuming, as it necessitates repeated scans of the sequence data and the search for candidate episodes. In this paper, we introduce a novel approach for episode mining in a data stream. Our method distinguishes itself by scanning newly added data to update existing frequent episodes, all without the need for scanning the original data or searching for candidate episodes. The experiments also show that our approach is more efficient compared to other existing methods.