Predicting web development effort using a bayesian network

摘要: OBJECTIVE - The objective of this paper is to investigate the use a Bayesian Network (BN) for Web effort estimation. METHOD We built BN automatically using HUGIN tool and data on 120 projects from Tukutuku database. In addition model node probability tables were also validated by project manager well-established company in Rio de Janeiro (Brazil). accuracy was measured 30 (validation set), point estimates (1-fold cross-validation 80%-20% split). obtained compared forward stepwise regression (SWR) as one most frequently used techniques software estimation. RESULTS Our results showed that BN-based predictions better than previous Web-based cross-company models, significantly SWR. CONCLUSIONS suggest that, at least dataset used, allows representation uncertainty, inherent estimation, can outperform other commonly such those multivariate techniques.