# commitmessagequality

**Repository Path**: ecust-dp/commitmessagequality

## Basic Information

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

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2024-07-02
- **Last Updated**: 2024-09-16

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Commit Message Quality
Code replication for **Commit Message Quality** experiments in the *ICSE 2023* paper: 
Commit Message Matters: Investigating Impact and Evolution of Commit Message Quality
&the *ICSE 2022* paper: 
What Makes a Good Commit Message?

## python environment setup:
```shell
conda create -n commit_message_quality python=3.8
pip install allennlp allennlp-models
```

## Three Part of Experiments:
### Part 1: preprocess commit message
1. clone the **project** in **cloned** folder
2. prepare the **data.csv** include **commit_hash**
3. run the **preprocess.py** by:
```shell
python preprocessor.py {current_project_commitid_csv} {project_name} {github token} {orgnazition} {marker}
```



### Part 2: train and evaluate the model


### Part 3：evaluate your commit message quality
1. put the dataset with the py file in the same folder
2. run on the GPU machine
3. move the result to the **model_predict** folder
4. run the **vote.ipynb**


## Data exxample for JIT-Fine(JIT-DEFECT4J) Datasets
- Preprocess_data_newest
    - which is the newest preprocessed commit message data extracted in 2024/09/10
- JITFine_unlabel_cmsgquality.csv
    - the unlabeled preprocessed commit message data for JIT-Fine
- model——predict
    - the model（XGBOOST,BIGRU,BILSTM） prediction result

### Overview of JIT-DEFECT4J Datasets Commit Message Quality
<center class="half">
<img src="./1.png" width="600" height="400" /><img src="./1-2.png" width="200" height="400" />
</center>