mysqlsc/sequential_recommendation

Sequence Models

Popularity Method

recommend items based on the number of occurrences

BPR_MatrixFactorization

Bayesian Personalized Ranking

FPMC

use Markov Chain、personalized recommendation

HRM

use max-pooling、mean-pooling to generate session representation

GRU4Rec

use GRU for every session. There is no connection between sessions.

NARM

use global encoder and local encoder for single session.

HGRU

use user embedding、GRU for every session. There is a connection between each person's session.

SHAN

use pooling and attention to model all users' record

STAMP

use pooling、attention、MLP to model single session

SR-GNN

build graph、use GNN

reference

1、Rendle, Steffen , et al. "BPR: Bayesian Personalized Ranking from Implicit Feedback." Conference on Uncertainty in Artificial Intelligence AUAI Press, 2009.

2、Rendle, S., Freudenthaler, C., & Schmidt-Thieme, L. (2010). Factorizing personalized Markov chains for next-basket recommendation. Proceedings of the 19th International Conference on World Wide Web.

3、Wang, Pengfei, et al. "Learning Hierarchical Representation Model for NextBasket Recommendation." (2015):403-412.

4、Session-based Recommendations with Recurrent Neural Networks,2016,ICLR.

5、Jing Li, Pengjie Ren, Zhumin Chen, Zhaochun Ren, Tao Lian and Jun Ma (2017). Neural Attentive Session-based Recommendation. In Proceedings of CIKM'17, Singapore, Singapore, Nov 06-10, 2017.

6、HGRU2Rec: Personalizing Session-based Recommendation with Hierarchical Recurrent Neural Networks,2017,RecSys.

7、Sequential Recommender System based on Hierarchical Attention Network,2018,IJCAI.

8、STAMP: Short-Term Attention/Memory Priority Model for Session-based Recommendation,2018,KDD.

9、SR-GNN: Session-based Recommendation with Graph Neural Networks,2019,AAAI.