[4L3-GS-4] Web intelligence

[4L3-GS-4-02] Proposal on Generation of Rating Matrix Based on Rational Behavior on Online Recipe Site

〇Atsushi Yoshimoto¹, Hiroki Shibata¹, Yasufumi Takama¹ (1. Tokyo Metropolitan University)

Keywords:recommendation, collaborative filtering, recipe recommendation

This paper proposes a method for generating an artificial rating matrix that is to be used in collaborative filtering, assuming rational behavior of users in recipe selection. While collaborative filtering can recommend items using rating matrix only, it is sometimes difficult to collect a rating matrix with enough size. The proposed method generates a rating matrix by extracting features from an actual recipe dataset, which are used to simulate the rational behavior of users selecting recipes to cook. The experimental results show that it is possible to generate artificial rating matrix that is easier to predict than actual rating matrix and that considering randomness can bring the recommendation accuracy closer to that of actual rating matrix.