Mining High Utility Patterns under Guidance

Xin Wang

doi:10.1109/tbdata.2026.3718742

Mining High Utility Patterns under Guidance

2026-07-30

0

PRE

AI

DOI：10.1109/tbdata.2026.3718742

Original

Original request for help

Save

Abstract

En 中文

Frequent Pattern Mining (FPM) plays a crucial role in, e.g., social analytics. However, FPM faces the challenge of striking a balance between the computational cost and results quality. To address the issue, we investigate the problem of high utility patterns mining under guidance and propose a comprehensive approach to the issue. We first introduce a utility function, that integrates the support metric with pattern size to assess the utility value of a pattern. We then develop a distributed mining algorithm GIMINER, that relies on a guidance network to mine high utility patterns. In particular, the guidance network, that is built up through calculating prior probability among structurally related patterns on a (small sample) graph guides the pattern expansion. By incorporating time constraint, GIMINER tackles the FPM task in a flexible way, well balancing the computational cost and result quality. As verified via experimental studies, our algorithms are of superior effectiveness as well as efficiency, compared with existing state-of-the-art methods on real-life graphs.

Keywords:

Frequent Pattern Mining

Pattern Utility

Distributed Pattern Mining

Guidance Network

Mining High Utility Patterns under Guidance

Abstract

Journal

Organization

Cited Papers