SBMx

Alleviating Popularity Bias in Session-based Recommendation Considering Long-tail Distribution

Overall Framework of SBMx

Setups

Datasets

The dataset name must be specified in the --dataset argument

• Yoochoose 1/64 - using latest 1/64 fraction due to the amount of full dataset
• Diginetica
• RetailRocket

After downloaded the datasets, you can put them in the folder Datasets/ and preprocess datasets by running Datasets/preprocess_code/{dataset_name}.ipynb below line.

Train and Test

python main.py \
    --dataset diginetica \
    --batchSize 128 \
    --hiddenSize 100 \
    --epoch 30 \
    --lr 0.001 \
    --lr_dc 0.1 \
    --lr_dc_step 3 \
    --l2 1e-5 \
    --step 1 \
    --mixup_lam 0.9 \
    --mixup_pct 0.5 \

Citation

Please cite our paper if you use our code:

Heeyoon Yang, Jee-Hyong Lee.(2022).
Alleviating Popularity Bias in Session-based Recommendation Considering Long-tail Distribution.
한국정보과학회 학술발표논문집,(),532-534.