Abstract
Quantitative project management has an important role in software development projects which become increasingly massive and unmanageable. However, traditional approaches such as multiple regression analysis and Rayleigh model have difficulties in dealing with actual software development data which often involves low accuracy data, outliers, and missing values. In this paper, we propose a new approach using project failure risk prediction model based on Naive Bayes classifier. The proposed method is good at dealing with low accuracy data and missing values involved in software development data, and can provide incremental predictions in accordance with data acquisition during project execution.
