Knowledge-Graph-4-VIS-Recommendation

The implementation of paper "KG4Vis: A Knowledge Graph Based Approach for Visualization Recommendation". For more details related to this project, please visit our project page.

Instruction

1. Before running it, please download the raw data here and extract the .csv file to ./data.
2. Extract features: python feature_extraction.py under ./feature_extraction. We also provide the extracted feature to save time. Please download here and extract the .csv file to ./features.
3. KG construction and test generation: python KG_construction.py and python test_generation.py under ./KG_construction.
4. Embedding learning: ./run.sh to run embedding learning under ./embedding_learning. To tune parameters, please modify the shell file according to the possible parameters described in ./embedding_learning/codes/run.py.
5. Inference and rule generation: python inference.py and python rule_generation.py under ./inference. The results of inference is saved under ./inference_results. (The core algorithms of inference.py and rule_generation.py are the same. To faciltate easy usage, we create two files.)

Since it is still an experimental version, please feel free to let us know if there is any issue.

Key Package

Name	Version
python	3.7.9
scikit-learn	0.21.0
numpy	1.16.3
editdistance	0.5.3
pandas	0.24.2
pytorch	1.7.0

Credit

We would like to thank Dr. Kevin Hu for granting us the permission of using and open-sourcing the code and dataset in VizML.

Partial implementation is based on:

• https://github.com/mitmedialab/vizml
• https://github.com/guillermo-navas-palencia/optbinning
• https://github.com/DeepGraphLearning/KnowledgeGraphEmbedding