A Novel Classifier for a Kansei Recommender System

摘要

We propose a novel classifier for a Recommender System which is based on a Kansei Model in this paper. We called this Recommender System as Kansei Recommender System (hereafter, we denoted as KRS algorithm). The purpose of building KRS algorithm is to reduce the time of training data from database and give more precise recommender items for consumers by considering their Kansei (a Japanese word which means the consumers' psychological feeling). To build a novel classifier, we divide the KRS algorithm into two parts of algorithms: (1) Algorithm 1 is proposed to extract Kansei factors (score 1) and evaluation factors (score 2) from consumers' shopping items. (2) Algorithm 2 is proposed to give a training dataset that is to fit the scored value of Kansei model. Combining two algorithms, we get a novel classifier for a KRS algorithm. We give an architecture of KRS algorithm based on the database of on-line shopping market in the end of this paper.