Data Augmentation Using User Attention for Educational Content Recommendation Based on FFM
Kazuma Ohtomo, Ryosuke Harakawa, Masaki Iisaka, Masahiro Iwahashi · 2020
This paper proposes a method for augmenting evaluation data using each user's attention (view counts for each attribute of educational contents) for Field-aware Factorization Machines (FFM). In this study, we assume that the target user has no evaluation data. In the proposed method, by using the target user's attention, we retrieve a similar user from active users. Then, the evaluation data of the active user is duplicated for the target user. By training FFM using the duplicated evaluation data, personalized recommendation for the target user becomes feasible. Experimental results using educational contents on Forestanet show the effectiveness of the proposed method.