• IMRec-TF2
• • Set Up
  • Dataset
  • • Taobao-500K
  • Model Training
  • Key Arguments
  • Model Checkpoint

IMRec-TF2

This is the TensorFlow 2 implementation for the paper:
An Intention-aware Markov Chain based Method for Top-K Recommendation (link)

Set Up

Run

pip install -r requirements.txt

Dataset

Taobao-500K

It is a sample with around 500,000 interactions from the Taobao dataset (link). The dataset will be automatically generated for the first time you train the model. For preprocessing details, see utils.py.

Model Training

To train our model in the Taobao-500K dataset, please run

python train.py --dataset taobao --mode train_500K --model IMRec

Experimental settings are defined in config.py. For easy use, some of the settings can be overwritten by passing command-line arguments. For example, to set the embedding dimension as 50, run

python train.py --dataset taobao --mode train_500K --model IMRec --embed_dim 50

The model is evaluated on the test datasets automatically at the end of each epoch. Check the experimental results in the log folder.

Key Arguments

Some key arguments:

embed_dim: the embedding dimension

maxlen: the max length of interaction sequence

att_len: the order of the Markov chain

alpha: the long-term and short-term preference weighting factor

without_il: whether to disable the intention loss module

bpr: whether to use BPR loss

time_threshold: the time threshold of strict short-term definition. 0 for not setting the threshold

item_intention: whether to use (item, action) as intention

Model Checkpoint

We provide the best model of the Taobao-500K dataset. To load the model, run:

python train.py --dataset taobao --mode train_500K --model IMRec --weights_dir ./best_model/best_epoch.ckpt

If you want to save checkpoints when training, run:

python train.py --dataset taobao --mode train_500K --model IMRec --save_weights 1