"JITBoost" stands at the intersection of artificial intelligence and software defect prediction. In an era dominated by expansive software architectures, our initiative introduces an ensemble approach by integrating Boolean Combination Algorithms (BCAs) with traditional Machine Learning (ML) classifiers. This amalgamation aims to refine the accuracy and efficiency of defect prediction within vast execution traces.
The ever-growing complexity of software systems brings forth enormous execution traces, making the early identification of software anomalies paramount. Our solution leverages ensemble techniques, combining the strengths of multiple classifiers to address this challenge head-on.
- BBC2: Pair-wise Brute-force Boolean Combination
- IBC: Iterative Boolean Combination
- PBC2: Pruning Boolean Combination
- WPBC2: Weighted Pruning Boolean Combination
- WPIBC: Weighted Iterative Boolean Combination
- Random Forest
- Decision Tree
- Logistic Regression
- Naive Bayes
- KNN
- SVM
Dive deep into our organized repository to navigate through datasets, algorithms, and results:
- Metrics:
- CK
- CK_NET
- CK_PROC
- NET
- NET_PROC
- CK_NET_PROC
- Each metric is complemented by 27 unique datasets.
- Comprehensive dataframes displaying our ensemble's performance.
- In-depth visual Receiver Operating Characteristics (ROC) curves.
- Scripts for:
- Cross-validation
- ML classifier training.
BBC2_Algorithm: Python script detailing the BBC2 logic.IBC_Algorithm: Delve into the IBC technique.PBC2_Algorithm: Explore the PBC approach.WPBC2_Algorithm: Script for WPBC2 logic.WPIBC_Algorithms: Breakdown of the WPIBC method.
Peruse individual directories for a deeper understanding of each algorithm and dataset.
- Authors: Mohammed A. Shehab, Prof. Abdelwahab Hamou-Lhadj and Venkata Sai Gunda
- Affiliations: Concordia University & IIT Kharagpur
For inquiries, feedback, or partnership opportunities:
Professor Wahab Hamou-Lhadj 📧 Email: wahab.hamou-lhadj@concordia.ca