README.md

metadata

annotations_creators:
  - expert-generated
language:
  - en
license: cc-by-4.0
pretty_name: Cross-Project Defect Prediction (CPDP) Dataset — C & Java Projects
size_categories:
  - 1M<n<10M
source_datasets:
  - original
tags:
  - source-code
  - software-engineering
  - defect-prediction
  - transfer-learning
  - static-analysis
  - c-language
  - java
task_categories:
  - tabular-classification
task_ids:
  - multi-class-classification

🧩 Cross-Project Defect Prediction (CPDP) Dataset — C & Java Projects

This repository hosts a custom dataset for Cross-Project Defect Prediction (CPDP) research, curated from a diverse collection of real-world open-source projects written in C (441 projects) and Java (98 projects).
The dataset aims to support research on software defect prediction, transfer learning, and imbalanced data handling across heterogeneous programming environments.

---

📘 Overview

Language	#Projects	Description
C	441	Includes diverse open-source repositories from various domains
Java	98	Covers projects from academic domains collected from GitHub and other public repositories

Each project folder typically includes:

• Source code files (.c, .h, .java)
• Bug/defect labels (where available)
• Metadata (e.g., LOC, complexity, commits)
• Preprocessed CSV feature files for ML models

---

🎯 Purpose

The dataset is designed for:

• Cross-Project Defect Prediction (CPDP)
• Transfer Learning and Domain Adaptation studies
• Feature engineering on static code metrics
• Benchmarking new software defect prediction models

---

🧠 Suggested Research Directions

• Comparison of within-project vs cross-project prediction accuracy
• Study of language heterogeneity in CPDP (C ↔ Java transfer)
• Use of oversampling or class balancing methods (e.g., SMOTE, OTOMO)
• Integration with Bayesian Networks, Deep Learning, or Tensor-based models