# CRKM

**Repository Path**: kyle-liao/crkm

## Basic Information

- **Project Name**: CRKM
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 2
- **Forks**: 0
- **Created**: 2023-12-01
- **Last Updated**: 2025-02-17

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

* # CRKM
https://gitee.com/kyle-liao/crkm
  ## Environment Requirements

  - Ubuntu Operating System
  - Python >= 3.8
  - PyTorch 1.7+
  - Nvidia GPU with CUDA 11.1+ support

  ## Datasets

  We conduct experiments on three popular datasets: Amazon-book, Last-FM, and Yelp2018.

  - To construct the cold start scenario, we identify user registration time, item publishing time, or the first interaction time in the complete recommendation dataset. Subsequently, we partition the data chronologically.
  - For Amazon-book, download the book reviews (5-core) from [here](http://jmcauley.ucsd.edu/data/amazon) and place it in the corresponding `rawdata` folder.
  - For Last-FM, download the LFM-1b dataset from [here](http://www.cp.jku.at/datasets/LFM-1b/), extract it, and place it in the relevant `rawdata` folder.
  - For Yelp2018, download the dataset from [here](https://www.heywhale.com/mw/dataset/5ecbc342fac16e0036ec41a0), extract it, and place it in the corresponding `rawdata` folder.

  ## Negative Label Data Processing

  1. We provide the cold start scenario data for Last-FM. The code to construct data for other datasets is as follows.

     ```
     python construct_data.py
     ```

  2. Label processing

     ```
     python ./utility/label_data.py
     ```

  3. Node path processing

     ```
     python ./utility/make_path_list.py
     ```

  4. Data negative label processing

     ```
     python ./utility/negative_label.py
     ```

  ## Training

  1. Here, we provide the "meta-model" after meta-training, so you can directly adapt to the cold start scenario.

     ```
     python main.py --dataset last-fm --use_meta_model True
     ```

  2. Alternatively, you can retrain the entire model.

     ```
     python main.py --dataset last-fm --use_meta_model False
     ```

  ## References

  - We used the code from [MetaKG](https://github.com/ZJU-DAILY/MetaKG).
  - You can find other benchmark models on GitHub.