[4K3-GS-10] AI application: Movement

[4K3-GS-10-05] Data Fusion with Limited Shared Information

〇Masataka Ushiku¹, Yusuke Kumagae¹, Yokoi Sho^2,3 (1. Hakuhodo DY Holdings Inc., 2. Tohoku University, 3. RIKEN)

Keywords:Data Fusion, Optimal Transport

Data fusion is a methodology aimed at integrating multiple datasets collected independently, with the goal of transforming them into a unified dataset for analysis. In this paper, we address situations where a limited number of attributes such as age or gender, which we call Shared Information, are observed across domains without common users or other common features. We propose a method to estimate similarity between users that leverages this shared information as clues when utilizing an optimal transport algorithm. The effectiveness of our proposed method is demonstrated through numerical experiment using multiple real datasets.