# JITLine

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

## Basic Information

- **Project Name**: JITLine
- **Description**: Code replication for JIT-DP and DL experiments in the MSR 2021 paper: JITLine: A Simpler, Better, Faster, Finer-grained Just-In-Time Defect Prediction
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2025-06-04
- **Last Updated**: 2025-06-11

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# JITLine

#### 1. Project Description
Code replication for JIT-DP and DL experiments in the MSR 2021 paper: JITLine: A Simpler, Better, Faster, Finer-grained Just-In-Time Defect Prediction

#### 2. System Environment

**P40-1**_高性能计算平台应用名称：gym

    CPU：10 核
    
    RAM：100 GB 
    
    GPU：NVIDIA Tesla P40 24G
    
    OS：Ubuntu 18.04

#### 3. Environment Settings

```git clone https://gitee.com/ecust-dp/jitline```

```cd jitline/JITLine```

```conda create --name JITLine python=3.9.1```

```conda activate JITLine```

```pip install -r requirements.txt```

**Note:** The Pro edition of PyCharm is needed for running Jupyter Notebook

#### 4. Data Preparation

Manually download **JITDefectPrediction.tar.gz** from 高性能计算平台(Path: ~/ECUST-SE/DL/JITLine/), upload it to the **JITLine/** folder and extract the **data** folder containing **openstack_train/test/dict.pkl and qt_train/test/dict.pkl** via ```tar -zxvf data.tar.gz```

**Note:** These data files are obtained from the **data+model/data/jit/** folder in the [Zenodo-data](https://zenodo.org/record/3965149#.X2VeP5MzY1J) shared in the [GitHub-CC2Vec](https://github.com/CC2Vec/CC2Vec) 

Table 2: The statistics of OpenStack and Qt.

| Project   | # Commits | % Defective | # Unique Tokens | Avg. Commit Size | Avg. Defective Lines | 
|-----------|-----------|-------------|-----------------|------------------|----------------------|
| OpenStack | 12,374    | 13%         | 32K             | 73 LOC           | 53%                  |
| Qt        | 25,150    | 8%          | 81K             | 140 LOC          | 51%                  |


#### 5. Run JITLine

**RQ1-3**

Follow the steps in **JITLine_RQ1-RQ3.ipynb**

**RQ4**

Follow the steps in **JITLine_RQ4.ipynb**

#### 6. Replication Results

**(RQ1) The evaluation result of our JITLine approach compared with the state-of-the-art approaches for Just-In-Time defect prediction.**
![img.png](RQ1.png)

**(RQ2) The cost-effectiveness of our JITLine approach compared to the state-of-the-art approaches for Just-In-Time defect prediction with respect to PCI@20%Recall, Effort@20%Recall, and POpt.**
![img.png](RQ2.png)

**(RQ3) The average CPU and GPU computational time (minutes±95% Confidence Interval) of the model train ing of JIT defect prediction approaches after repeating the  experiment 5 times.**
![img.png](RQ3.png)

**(RQ4) The results of our JITLine at the line level when compared to the N-gram-based line-level JIT defect prediction approach of Yan et al. with respect to Top-10 Accuracy(↗), Recall@20%Effort(↗), Effort@20%Recall(↘), and IFA(↘). The higher (↗)or the lower (↘) the values are, the better the approach is.**
![img.png](RQ4_Paper.png)
![img.png](RQ4_P40-1.png)
![img.png](RQ4.png)