Proposal of a Swarm Intelligence-Based Pattern Mining Method Utilizing the Locality of Patterns

Abstract

In the modern era, where vast amounts of data are accumulated, enhancing the efficiency of pattern mining to extract valuable information from data is a critical challenge. Although pattern mining targets various types of data, patterns can localize on data when there is an order relationship. Traditional methods have not fully utilized this localization of patterns for reducing and optimizing the search space. This study aims to rapidly extract patterns localized on data by leveraging their localization. The proposed method employs swarm intelligence through agents that perform local pattern synthesis and search, enabling efficient extraction by utilizing the patterns' localization. Experiments comparing the accuracy and speed of localized pattern extraction with existing methods confirmed that the proposed method can extract localized patterns more rapidly while maintaining high extraction accuracy.