SIMS

This work "Modeling User Interests With Online Social Network Influence by Memory Augmented Sequence Learning" has been published in TNSE 2021.

📄 Description

SIMS is a novel social-based sequence learning model for predicting the types of items/PoIs that a user will likely buy/visit next. Specifically, SIMS leverages the sequence-to-sequence learning method to learn a representation for each user sequence. Moreover, an autoencoder-based model was proposed to learn social influence, which is integrated into SIMS for predicting user interests. In addition, SIMS employs DNC to further improve prediction accuracy.

🔧 Dependencies

• Python == 3.7 (Recommend to use Anaconda or Miniconda)
• PyTorch == 1.8.1
• NVIDIA GPU (RTX 3090) + CUDA 11.1

Installation

1. Clone repo

   git clone https://github.com/BIT-MCS/SIMS.git
   cd SIMS

2. Install dependent packages

   pip install -r requirements.txt
   

💻 Training

We provide complete training codes for SIMS.
You could adapt it to your own needs.

1. If you don't have NVIDIA RTX 3090, you should comment these two lines in file SIMS/main.py.

   [17]  torch.backends.cuda.matmul.allow_tf32 = False
   [18]  torch.backends.cudnn.allow_tf32 = False
   

2. You can modify the config file SIMS/conf.py for model training.
   For example, you can control the size of RNN in the model by modifying this line

   [8]  'rnn_size': 128,
   

3. Training

   cd SIMS
   python main.py
   

   The log files will be stored in [SIMS/log].

🏁 Testing

1. Testing

   cd SIMS
   python test.py
   

📧 Contact

If you have any question, please email [email protected].

Paper

If you are interested in our work, please cite our paper as

@ARTICLE{9294053,
  author={Wang, Yu and Piao, Chengzhe and Liu, Chi Harold and Zhou, Chijin and Tang, Jian},
  journal={IEEE Transactions on Network Science and Engineering}, 
  title={Modeling User Interests With Online Social Network Influence by Memory Augmented Sequence Learning}, 
  year={2021},
  volume={8},
  number={1},
  pages={541-554},
  doi={10.1109/TNSE.2020.3044964}
}