/BAFLineDP

Primary LanguagePython

BAFLineDP: Code Bilinear Attention Fusion Framework for Line-level Defect Prediction

This paper has been accepted in proceedings of the 31st international conference on software analysis, evolution and reengineering (SANER 2024).

Environment Setup

Python Environment Setup

The implementation codes are running in the following environment setups.

  • python == 3.9.13
  • pytorch == 1.12.1
  • numpy == 1.23.2
  • pandas == 1.5.3
  • transformers == 4.20.1
  • joblib == 1.2.0
  • more-itertools == 9.1.0
  • rdkit == 2023.3.1
  • scikit-learn == 1.1.2
  • tqdm == 4.64.1

R Environment Setup

Download the following packages: tidyverse, gridExtra, ModelMetrics, caret, reshape2, pROC, effsize, and progress.

Experiment

Experimental Setup

The following parameters are used to train our BAFLineDP model.

  • batch_size = 16
  • num_epochs = 10
  • embed_dim = 768
  • gru_hidden_dim = 64
  • gru_num_layers = 1
  • bafn_hidden_dim = 256
  • dropout = 0.2
  • lr (learning rate) = 0.001

Code Preprocessing

Download the datasets from the github_url and keep them in datasets/original/.

Run the following command to prepare data for model training. The output will be stored in datasets/preprocessed_data/.

python code_preprocessing.py

BAFNLineDP Model Training

To train BAFNLineDP model for each project, run the following command. The trained models will be saved in output/model/BAFLineDP/, and the loss will be saved in output/loss/BAFLineDP/.

python train_model.py

Prediction Generation

To make a prediction within software projects, run the following command.

python generate_within_prediction.py

The generated outputs of within-prediction are stored in output/prediction/BAFLineDP/within-release/.

To make a prediction across software projects, run the following command.

python generate_cross_prediction.py

The generated outputs of cross-prediction are stored in output/prediction/BAFLineDP/cross-release/.

Get the Evaluation Results

To obtain file-level and line-level defect prediction results within WPDP and CPDP scenarios, you need to set the absolute working path for the script to run first, and then execute the following command.

Rscript  get_results.R

The results are stored in output/result/BAFLineDP/.