Parallel frequent itemset mining on streaming data

Abstract

Owing to the widely used of data stream, frequent itemset mining on data stream have received more attention. Data stream is fast changing, massive, and potentially infinite. Therefore, we have to establish new data structure and algorithm to mine it. On the base of our previous work, we propose a new paralleled frequent itemset mining algorithm for data stream based on sliding window, which is called PFIMSD. The algorithm compresses whole data in current window into PSD-trees on paralleled processor only by one-scan. Increment method is used to append or delete related branch on PSD-tree when window is sliding. The experiment shows PFIMSD algorithm has good performance on efficiency and expansibility.