Enhancing Healthcare Recommendations: A Privacy-Protective and Interpretable Cross-Domain Framework
Authors: Xun Liang, Zhiying Li, Hongxun Jiang
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The model s effectiveness and scalability were validated using three public datasets and a healthcare cross-domain recommendation dataset. In addition to traditional evaluation metrics, strong privacy metrics and the unique sentence ratio were used to assess privacy protection and interpretability. We also compared the characteristics of privacy protection and interpretability between e-commerce and healthcare recommendation scenarios. This section aims to answer the following research questions through experiments and case studies. Q1: Can the proposed HCR model achieve better performance and enhanced privacy compared to existing SOTA models of plaintext CDR, multimodal CDR, and advanced PPCDR? Q2: Can the SS (Self-Supervised modality-aware encoder) and PP (Privacy Protection) submodules within HCR enhance its performance? Q3: In the field of healthcare services, does the explanations provided by HCR significantly enhance patients understanding of the recommendations? Q4: How do various hyperparameters affect the performance of HCR? |
| Researcher Affiliation | Academia | Xun Liang, Zhiying Li, Hongxun Jiang* School of Information, Renmin University of China No. 59 Zhongguancun Street, Beijing, 100872, P.R. China EMAIL |
| Pseudocode | No | The paper describes its methodology using mathematical formulations (Eqs. 1-18) and descriptive text, but it does not contain a clearly labeled 'Pseudocode' or 'Algorithm' block, nor does it present structured steps formatted like code. |
| Open Source Code | Yes | Code and datasets https://github.com/zyl-mc/HCR |
| Open Datasets | Yes | The model s effectiveness and scalability were validated using three public datasets and a healthcare cross-domain recommendation dataset. This paper first compares the performance of our proposed HCR method with advanced benchmark algorithms through extensive experiments on a large-scale public Amazon dataset (Cao et al. 2022). We also collected a dataset for online medical consultation (OMC) services... All data used is publicly available and collected compliantly. Viewer IDs are anonymized, depersonalized codes are used, and private information is manually excluded. |
| Dataset Splits | No | The paper states that "Early stopping and validation follow Light GCN s approach" and mentions evaluating performance by "cutting off the ranked list at 5", but it does not provide specific details on how the datasets were split into training, validation, and test sets (e.g., percentages, sample counts, or explicit splitting methodology). |
| Hardware Specification | No | The paper does not provide specific hardware details such as GPU models, CPU types, or memory amounts used for running the experiments. It only mentions software like PyTorch and a pretrained GPT-2 model. |
| Software Dependencies | No | Our HCR model is implemented in Py Torch with key parameters tuned for optimal performance. For explanations, we use a pretrained GPT-2 model from huggingface with Byte Pair Encoding (BPE) to handle rare words. The paper mentions 'Py Torch' and 'GPT-2' but does not provide specific version numbers for these software components. |
| Experiment Setup | Yes | We use the Adam optimizer with a learning rate of 0.001 and a regularization coefficient of 0.0001. Early stopping and validation follow Light GCN s approach. For the self-supervised task, we set the temperature τss = 0.5. In the privacy-preserving synthesizer, we set a fixed λpp = 0.4. For explanations, we use a pretrained GPT-2 model from huggingface with Byte Pair Encoding (BPE) to handle rare words, setting the length to 20 BPE tokens and embedding size to 768. |