| THONGTANUNAM PATANAMON | 1461020: D, 中間発表 | ソフトウェア設計学 飯田 元, 松本 健一, 市川 昊平, 渡場 康弘 |
|
Title: Studying Review Involvement in Modern Code Review Process
Abstract: Software code review is a well-established software quality practice. Recently, Modern Code Review (MCR), a lightweight variant of the code review process, has been widely adopted in both open source and proprietary projects. Due to human-intensive nature of code reviewing, review involvement plays a key role in code review effectiveness. However, the MCR process may suffer from lax review involvement, since it often lacks mechanisms for ensuring a base level of review involvement. Hence, in this research, we set out to investigate the impact that review involvement has on software quality. In particular, we investigate the review involvement in terms of (1) review activity that has been applied during MCR reviews; and (2) reviewing expertise of reviewers. Through case studies of open source projects, we demonstrate the nature of lax review involvement and its impact on software quality. Our results highlight the need for reviewing policies to mitigate the risk of having post-release defects in software products. We believe that our results and empirical observations help to support the management of the MCR process and the adherence to our recommendations will lead to a more effective review process. Language of the presentation: English | ||
| TANTITHAMTHAVORN CHAKKRIT | 1461019: D, 中間発表 | ソフトウェア工学 松本 健一, 飯田 元, 門田 暁人, 伊原 彰紀, 畑 秀明 |
|
Title: Towards a Better Understanding of Experimental Components on Defect Prediction Modelling
Abstract: Defect prediction models are used to pinpoint risky software modules and understand past pitfalls that lead to defective modules. The predictions and insights that are derived from defect prediction models may not be accurate and reliable if practitioners do not consider the impact of experimental components (e.g., datasets, metrics, and classifiers) of defect prediction modelling. Therefore, a lack of awareness and practical guidelines from previous research can lead to invalid predictions and unreliable insights. In this thesis, we investigate the impact that experimental components (i.e., realistic noise in defect datasets, classifier parameters, and model validation techniques) have on the performance and interpretation of defect prediction models. Through case studies of systems that span both proprietary and open-source domains, we demonstrate that (1) realistic noise in defect datasets is considered harmless, suggesting that practitioners can rely on defect labelling of defect datasets; (2) parameter settings of classification techniques substantially change the performance and interpretation of defect prediction models, suggesting that practitioners should apply automated parameter optimization; and (3) the out-of-sample bootstrap validation produces the best trade-offs between bias and variance of performance estimates, suggesting that the single holdout and cross-validation families should be avoided. These observations leads us to conclude that practitioners should carefully select experimental components in order to produce more accurate and reliable defect prediction models. Language of the presentation: English | ||
| GEORGESCU LIVIU MARIUS | 1461208: D, 中間発表 | インターネット工学 山口 英, 安本 慶一, 藤川 和利, 門林 雄基 |
|
Title: IPv6NET: A Suite of Methodologies for Evaluating IPv6 Transition Technologies
Abstract: Among the many challenges introduced by the IPv6 transition process to the Internet community, one of the most difficult was presented to the network operators. All of the existing production networks were forced to reconsider their inner architecture to move towards IPv6. To support network operators in this challenge, the IETF has proposed multiple IPv6 transition and coexistence technologies. Which one of these technologies is most feasible for their scenario, is still an open problem. To support network operators solve this problem, we are proposing a suite of practical evaluation methodologies, exploring various feasibility dimensions of transition technologies. The methodologies are associated with a heterogeneous IPv4 and IPv6 network testbed, which we called the IPv6 Network Evaluation Testbed (IPv6NET). In order to validate these methodologies, we have used them to analyze the feasibility of two open source transition implementations, covering multiple transition technologies. To that end, we are showing how network performance, scalability and operational capability data can be obtained, analyzed and compared. Language of the presentation: English | ||