nakhlarafi / gnet4fl

GNet4FL

GNet4FL: A fault localization method based on graph convolutional neural networks to predict potential faulty locations via static and dynamic features of source code. This repository includes data processing and model constructing.

Environment

You need to have the following library

• Javalang https://github.com/c2nes/javalang
• Pytorch V1.60 https://pytorch.org

  Not sure if other versions of pytorch are supported.

• Numpy
• Scikit-learn

Organization

The following is the organization of the repository.

│  ast_statement.py
│  data_load.py
│  main.py
│  models.py
│  README.md
│  utils.py
│  
├─MS
│  │  model.py

• ast_statement.py : #Constructing ASTs
• data_load.py: #Loading processed data
• main.py :#Main file for running the model
• models.py : #Mainly GraphSAGE models
• utils.py : #Some parameter settings

Run

1. Data processing
2. Set the project and version name, e.g.,project='Chart',version=1
3. run main.py

Due to many model parameters and the fact that their initialization is not fixed, the results may differ from our paper.