Until now, software engineering researchers have used Case-based Reasoning,
Neural Networks, Genetic Programming, Fuzzy Logic, Decision Trees, Naive
Bayes, Dempster-Shafer Networks, Artificial Immune Systems, and several
statistical methods to build a robust software fault prediction model. Some
researchers have applied different software metrics to build a better prediction
model, but recent papers [29] have shown that the prediction technique is much
more important than the chosen metric set. The use of public datasets for software
fault prediction studies is a critical issue. However, our recent systematic review
study has shown that only 30% of software fault prediction papers have used
public datasets [5].