FPS-Tree Algorithm to Find Top-k Closed Itemsets in Data Streams

Abstract

Frequent itemset mining has become a popular research area in the data mining community and has been applied in various areas over the last few years. There are two main technical hitches when searching for frequent itemsets. The first is to provide an appropriate minimum support value, and the second is the generation of a large number of association rules. In many cases, users are only interested in finding only top-k frequent itemsets with some defined threshold value. In this paper, we present an algorithm to mine top-k frequent closed itemsets from streaming data using a sliding window approach. A fast algorithm is proposed to find frequent closed itemsets with user-defined minimum and maximum lengths to reduce the number of frequent itemsets. Moreover, we have also incorporated a bitmap-based data structure to improve the performance by eliminating multiple scans. Different datasets have been used for experimentation and to benchmark the proposed technique and algorithm.