Pattern Recommendation in Task-oriented Applications: A Multi-Objective Perspective [Application Notes]

TL;DR: This work proposes an effective multi-objective pattern mining evolutionary algorithm for finding optimal pattern set, which does not need to specify the prior parameters min_sup, min_occ and m.

Abstract: Task-oriented pattern mining is to find the most popular and complete pattern for task-oriented applications such as goods match recommendation and print area recommendation. In these applications, the measure support is used to capture the popularity of patterns, while the measure occupancy is adopted to capture the completeness of patterns. Existing methods for mining task-oriented patterns usually combine these two measures as one measure for optimization, and require users to set the prior parameters such as the minimum support threshold min_sup, the minimum occupancy threshold min_occ and the relative importance preference l between support and occupancy. However, it is very difficult for users to set optimal values for these parameters especially when they do not have any prior knowledge in real applications. To overcome this challenge, we propose an evolutionary approach for pattern mining from a multi-objective perspective since support and occupancy are conflicting. Specifically, we first transform this pattern mining problem into a multi-objective optimization problem. Then we propose an effective multi-objective pattern mining evolutionary algorithm for finding optimal pattern set, which does not need to specify the prior parameters min_sup, min_occ and m. Finally, we select k best patterns from the obtained pattern set for final pattern recommendation. Experimental results on two real task-oriented applications, namely, goods match recommendation in Taobao and print area recommendation in SmartPrint, and several large synthetic datasets demonstrate the promising performance of the proposed method in terms of both effectiveness and efficiency.