We have reimplemented/modified each approach based on their implementations to fit our dataset and experimental setup.
The original paper implemented it in R, but as there was no publicly released implementation, we implemented it in Python.
- original paper: Defect Reduction Planning (Using TimeLIME)
It was originally implemented based on k-1, k, k+1 and had some parts conflicting with the content of the paper, so we modified it to fit our experiment.
The original paper mentions the use of BigML, but there was no code related to automation, so we performed automation based on the BigML API in mining_sqa_rules.py.