Official Implementation of FRec

SIGIR'24 Paper: Modeling User Fatigue for Sequential Recommendation

Based on Microsoft Recommender and Tensorflow 2.1.

Data

We provide Taobao dataset. The input data is organized as standard sequential input in recommender described here.

unzip data.zip

Train Model

• Our FRec: python run.py --model model --name trial
• CLSR: python run.py --model clsr --name trial
• SLiRec: python run.py --model slirec --name trial
  For FRec, important hyper-parameters include,
• num_cross_layers: The number of cross layers
• recent_k: Truncated length $T$
• num_interests: The number of interests $K$
• k_size: Kernel size $s$ in 1D convolution
• alpha: Weight of contrastive learning $\alpha$