license: mit
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
- split: validation
path: data/validation-*
dataset_info:
features:
- name: CommitHash
dtype: string
- name: NewPath
dtype: string
- name: Diff
dtype: string
- name: FaultInducingLabel
dtype: int64
splits:
- name: train
num_bytes: 116120391
num_examples: 207464
- name: test
num_bytes: 38526893
num_examples: 69155
- name: validation
num_bytes: 38411094
num_examples: 69155
download_size: 65354656
dataset_size: 193058378
Dataset Card for TechDebt
This dataset was generated from The Technical Debt Dataset created by Lenarduzzi, et al. and the citation is down below.
Dataset Details and Structure
The labels for the dataset were provided by the SZZ algorithm cited by the paper and matched to the diff in the commit where the technical debt was located. This diff was then cleaned to only include the lines of code added.
Bias, Risks, and Limitations
Beware of the data imbalance if you would like to use the dataset. Also, the queries used to extract this data are still being checked over to ensure correctness.
Recommendations
Changes are constantly being made to this dataset to make it better. Please be aware when you use it.
References
Valentina Lenarduzzi, Nyyti Saarimäki, Davide Taibi. The Technical Debt Dataset. Proceedings for the 15th Conference on Predictive Models and Data Analytics in Software Engineering. Brazil. 2019.