作者: Michael Klas , Adam Trendowicz , Yasushi Ishigai , Haruka Nakao
DOI: 10.1109/ESEM.2011.33
关键词:
摘要: Reliable predictions are essential for managing software projects with respect to cost and quality. Several studies have shown that hybrid prediction models combining causal Monte Carlo simulation especially successful in addressing the needs constraints of today's industry: They deal limited measurement data and, additionally, make use expert knowledge. Moreover, instead providing merely point estimates, they support handling estimation uncertainty, e.g., estimating probability falling below or exceeding a specific threshold. Although existing methods do well terms uncertainty information, we can show leave coming from imperfect modeling largely unaddressed. One consequences is probably provide over-confident estimates. This paper presents possible solution by integrating bootstrapping into methods. In order evaluate whether this does not only theoretically improve estimates but also has practical impact on quality results, evaluated an empirical study using more than sixty six different domains application areas. The results indicate currently used realistic be significantly improved proposed solution.