DSR

Introduction

Deep Social Recommendation (DSR) is a new recommendation framework tailored to knowledge graph-based personalized recommendation. DSR fully exploits both the social influence from potential friends and the collaborative influence from interactions for better embedding learning.

Environment Requirement

• Python == 3.7
• numpy == 1.16.2

Datasets

• Ciao & Epinion

  • In the Ciao and Epinion datasets, we have user' ratings towards items. The data is saved in a txt file (rating.txt) and the format is as follows:

    userid itemid rating
    

  • trust.txt: it contains the trust relations between users. There are two columns and both of them are userid, denoting there is a social relation between two users.

    userid userid
    

  • train.txt: it contains data for train. Each line is a user with a list of her interacted items.

  • test.txt: it contains data for test. Each line is a user with a list of her test items.

Model

• Deep Social Collaborative Ranking (DSCR.py)

  • Model for Deep Social Collaborative Ranking

• Running Command

     python3 DSCR.py --dataset ciao --regs [1e-5] --embed_size 64 --layer_size [64,64,64] --layer_size_S [64,64] --lr 0.0005 --batch_size 1024 --epoch 400 
     
     python3 DSCR.py --dataset epinion --regs [1e-5] --embed_size 64 --layer_size [64,64,64] --layer_size_S [64,64] --lr 0.0005 --batch_size 1024 --epoch 400 
  

  You need to specify serveral parameters for training and testing:

  • dataset: ciao / epinion
  • regs: regularization weight
  • layer_size: the number of layers and embedding size for user-item interaction network
  • layer_size_s : the number of layers and embedding size for user-user social network
  • lr: learning rate
  • batch_size : the size of batch for training
  • epoch : the epoch for training