Lessons Learned from Using a Deep Tree-Based Model for Software Defect Prediction in Practice

Abstract

Defects are common in software systems and can potentially cause various problems to software users. Different methods have been developed to quickly predict the most likely locations of defects in large code bases. Most of them focus on designing features (e.g. complexity metrics) that correlate with potentially defective code. Those approaches however do not sufficiently capture the syntax and different levels of semantics of source code, an important capability for building accurate prediction models. In this paper, we develop a novel prediction model which is capable of automatically learning features for representing source code and using them for defect prediction. Our prediction system is built upon the powerful deep learning, tree-structured Long Short Term Memory network which directly matches with the Abstract Syntax Tree representation of source code. An evaluation on two datasets, one from open source projects contributed by Samsung and the other from the public PROMISE repository, demonstrates the effectiveness of our approach for both within-project and cross-project predictions. CCS CONCEPTS• Software and its engineering → Software creation and management; KEYWORDSSoftware engineering, software analytics, defect prediction ACM Reference Format:

μVulDeePecker: A Deep Learning-Based System for Multiclass Vulnerability Detection

“…A recent development is to exploit deep learning, which has a great potential in reducing the burden on human experts for defining features. VulDeePecker [13] is the first system using deep learning to detect vulnerabilities at slice-level, while noting that there are also related studies on using deep learning for vulnerability discovery at the function level [14], [51], defect prediction [15] and related tasks [16]. Above systems are binary classifiers by telling whether a piece of code is vulnerable or not.…”

“…These solutions reduce the workload on human experts because they only need to roughly define features for learning machine learning-based models that can detect vulnerabilities, rather than defining vulnerability patterns manually. Compared with traditional machine learning techniques, researchers have started using deep learning for detecting vulnerabilities [13], [14] and software defects [15], [16].…”

Actionable Analytics on Software Requirement Specifications

“…A method is suggested for estimating defectiveness of source code by Kapur and Sodhi [21]. Dam et al [22] introduced a deep tree-based model for software defect prediction.…”

Software Defect Prediction u sing Support Vectorised Data and Intelligent Techniques

“…Most of the software defect prediction techniques focus on complexity software metrics of the defective code. These techniques do not capture different levels of source code semantics which are vital elements for constructing a good prediction models [14]. So, Hoa Khanh Dam et al developed a novel prediction model that can learn the features that represent a source code automatically and can be used for prediction.…”

“…So, Hoa Khanh Dam et al developed a novel prediction model that can learn the features that represent a source code automatically and can be used for prediction. They applied Long Short Term Memory (LSTM) and deep learning techniques [14] for capturing features like structure of programs and their semantic information. Jian.…”