|
Title:
|
S-ISFMF: SOCIAL NETWORK AND IMAGE SHAPE
FEATURE MATRIX FACTORIZATION |
|
Author(s):
|
Takuya Tamada, Mingtian Gao and Ryosuke Saga |
|
ISBN:
|
978-989-8704-48-1 |
|
Editors:
|
Miguel Baptista Nunes, Pedro Isaías and Philip Powell |
|
Year:
|
2023 |
|
Edition:
|
Single |
|
Keywords:
|
Recommender System, Machine Learning, Matrix Factorization, Deep Learning |
|
Type:
|
Full Paper |
|
First Page:
|
153 |
|
Last Page:
|
162 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
Recommender system is a useful tool to help users to select the best item in today's information-overloaded society.
Probabilistic Matrix Factorization (PMF) is one of successful recommender system methods that introduces a probabilistic
algorithm to matrix factorization. Various auxiliary information can be used in PMF. However, there is no example of
using images and social networks at the same time. In this paper, we describe a model that extracts information from images
and social networks and incorporates it into PMF. In our method, we extracted contour information from item images and
optimized the user and item matrix in matrix factorization. We further optimized the user matrix using explicit and implicit
social networks among users. As a result, we achieved an accuracy improvement of 0.75%~7.25% on three real world
datasets compared to existing methods. |
|
|
|
|
|
|
