TopUMS: Top-k Utility Mining in Stream Data

Top-k utility mining has attracted great attention in the past few years. Its goal is to discover a set of patterns that have the k highest utilities in a database. Many algorithms have been proposed to efficiently discover top-k high utility itemsets (HUIs), but most of them assume that the database is static. The data are mostly streaming data with continuous, high-speed, and unrestricted features. Thus, it is not possible to store the complete stream data in the same manner as static data; real-time storage and processing are required. In this study, we propose a novel algorithm called TopUMS to mine the top-k HUIs in stream data using the sliding window model. Additionally, we propose a data structure called uList, which is constructed in a horizontal method without ineffective comparison operations. Moreover, we utilize the common batch utility between two consecutive windows to raise minimum utility threshold for the next sliding window. The experimental results demonstrate that TopUMS outperform the state-of-the-art algorithm in terms of execution time and memory.