Enhancing Personalized Recommender Systems through Explainable User Preference Mining and Reinforced Knowledge Graph Reasoning
The DualPGPR model is being reviewed.
The datasets used in the experiments are from Amazon available at http://jmcauley.ucsd.edu/data/amazon/. The preprocessed datasets are in the preprocess folder.
The required packages are listed in the requirements.txt file.
- Preprocess the datasets:
python preprocess.py --dataset <dataset_name>- Train the TransE embeddings:
python train_transe_model.py --dataset <dataset_name>Arguments:
--dataset: the name of the dataset (one of beauty, cloth, cell)--name: the name of the model--epochs: the number of epochs--batch_size: the batch size--lr: the learning rate--weight_decay: the weight decay--l2_lambda: the lambda for l2 regularization--max_grad_norm: the maximum gradient norm--embed_dim: the embedding dimension--num_neg_samples: the number of negative samples--steps_per_checkpoint: the steps per checkpoint
- Train RL agent:
python train_agent.pyArguments:
--dataset: the name of the dataset (one of beauty, cloth, cell)--max_acts: the maximum number of actions--max_path_len: the maximum number of hops--gamma: the discount factor--lr: the learning rate--batch_size: the batch size--epochs: the number of epochs--ent_weight: the weight for calculating the entropy loss--act_dropout: the action dropout rate--state_history: the number of historical states prior to the current state--hidden: the hidden size of the FC layer
- Test the model:
python test_agent.pyArguments:
--dataset: the name of the dataset (one of beauty, cloth, cell)--max_acts: the maximum number of actions--max_path_len: the maximum number of hops--run_path: whether to generate reasoning paths or not--run_eval: whether to run evaluation or not--get_neg_paths: whether to get negative paths or not, turn off if testing evaluation metrics only
[1] Yongfeng Zhang, Qingyao Ai, Xu Chen, W. Bruce Croft. "Joint Representation Learning for Top-N Recommendation with Heterogeneous Information Sources". In Proceedings of CIKM. 2017.
[2] Yikun Xian, Zuohui Fu, S. Muthukrishnan, Gerard de Melo, Yongfeng Zhang. "Reinforcement Knowledge Graph Reasoning for Explainable Recommendation." In Proceedings of SIGIR. 2019.
[3] Wang, X., Li, Q., Yu, D., Li, Q., & Xu, G. (2024). Reinforced path reasoning for counterfactual explainable recommendation. IEEE Transactions on Knowledge and Data Engineering.