Just-in-time defect prediction for mobile applications: using shallow or deep learning?

Abstract

Just-in-time defect prediction (JITDP) research is increasingly focused on program changes instead of complete program modules within the context of continuous integration and continuous testing paradigm. Traditional machine learning-based defect prediction models have been built since the early 2000s, and recently, deep learning-based models have been designed and implemented. While deep learning (DL) algorithms can provide state-of-the-art performance in many application domains, they should be carefully selected and designed for a software engineering problem. In this research, we evaluate the performance of traditional machine learning algorithms and data sampling techniques for JITDP problems and compare the model performance with the performance of a DL-based prediction model. Experimental results demonstrated that DL algorithms leveraging sampling methods perform significantly worse than the decision tree-based ensemble method. The XGBoost-based model appears to be 116 times faster than the multilayer perceptron-based (MLP) prediction model. This study indicates that DL-based models are not always the optimal solution for software defect prediction, and thus, shallow, traditional machine learning can be preferred because of better performance in terms of accuracy and time parameters.

Toward Reduction in False Positives Just-In-Time Software Defect Prediction Using Deep Reinforcement Learning

“…Commonly used standalone classifiers include Logistic Regression [3,23,24], Naïve bayes [25], Support Vector Machine [26], Decision Tree [27], and Neural Network [28]. Whereas for ensemblebased learners range from single ensemble learner such as Random Forest [29] to multi-layer ensembles [27,30,31]. Prior JIT-SDP has utilized machine learning with batch learning to formulate such prediction models.…”

Just-in-time software defect prediction based on feature selection

“…Huang, QG [5] and others proposed the MTL-DNN method based on multi-task learning to address the issue of insufficient labeled training data in both the source and target projects. Additionally, research by Raymon [6] suggested that, when comparing machine learning and deep learning on mobile platforms, machine learning exhibits better predictive performance. Zhu et al [7] .…”

Rand-Index Target Projective Gradient Deep Belief Network for Software Fault Prediction

“…With this type of design, mean absolute error rate was found to be comparatively less. Yet another deep learning based prediction method was proposed in [7] to focus on the accuracy and time aspect. Since recently deep learning techniques have achieved significant results in several areas of applications, there is a requirement to apply for all type of problems, i.e., software defect prediction.…”