Hybrid Recommendation Algorithms Based on ConvMF Deep Learning Model

DOI

10.2991/wcnme-19.2019.36How to use a DOI?

Keywords

recommender systems; ConvMF; DE-CNN; SDAE

Abstract

Due to ConvMF (Convolutional Matrix Factorization) use side information to improve the accurate of the prediction rating, it shows side information is important for rating prediction accuracy. but it does not make fully use of the features of the item description documents such as reviews, abstract, or synopses. To handle the problem, this paper proposes a novel model DE-ConvMF, which have double embedding layer in ConvMF, take more attention on the item side information. This double embeddings includes two part: one is general embedding layer, other is domain embedding layer, we combine general embedding with domain embedding as the embedding layers. Then we use Stack Donising Auto Encoder (SDAE) to deal with users side information(age, sex, occupation), Through the user ratings and labels to improve the accuracy of forecast scores. Extensive experiment results on movielens (ml-10M) datasets show that our new model outperforms other methods in effectively utilizing side information and achieves performance improvement.

Copyright

© 2019, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Jiakun Zhao
AU  - Zhen Liu
AU  - Huimin Chen
AU  - Jingbo Zhang
AU  - Qing Wen
PY  - 2019/06
DA  - 2019/06
TI  - Hybrid Recommendation Algorithms Based on ConvMF Deep Learning Model
BT  - Proceedings of the 2019 International Conference on Wireless Communication, Network and Multimedia Engineering (WCNME 2019)
PB  - Atlantis Press
SP  - 151
EP  - 154
SN  - 2352-538X
UR  - https://doi.org/10.2991/wcnme-19.2019.36
DO  - 10.2991/wcnme-19.2019.36
ID  - Zhao2019/06
ER  -