Research Of Bayesian Network Model In Software Project Risk Management

Software project risk management has become an increasing important field due to the increasingly pervasive role of software in today's world. To manage software project risk is to deal with software process uncertainty to a certain extent. Bayesian network is a favorable tool to model and resolve uncertainty. Bayesian network technique is believed as the new direction of software project risk management by domestic and overseas scholars. The focus of this dissertation is to establish a framework of software project risk management based on Bayesian network, and to study risk Bayesian network models, structure and probability inference method. Thereby, it provides a systematic modeling method for software project risk management, and it offers new technique approach and realization means for solving iterative and dynamic problem of software project risk. The main content and fruits of this paper are outlined as follows:In this dissertation, a risk management framework based on Bayesian network is put forward. The definition of risk Bayesian network is given. The general process and main management activities of this framework are presented.The two-stage modeling method for software project risk's static Bayesian networks is studied. The object-oriented modeling method for risk Bayesian network is presented, structure refinement method based on network measurement is discussed, and probability parameters refinement method based on Maximum Likelihood estimates is put forward. The two-stage modeling method takes into account the characteristics of software project risk management and software metrics data, integrates qualitative knowledge and quantitative data.To study the software project iterative process risk's Bayesian network model, the definition of Cyclic Bayesian network is presented, probability convergence property of directed cycle in Cyclic Bayesian network is proved and probability inference method is put forward. Cyclic Bayesian network provides modeling method and inference algorithms for the management of software project iterative process risk.To study software project time related risk in software project process, the definitions of Temporal Bayesian network and Scheduling Bayesian network are given. The probability updating algorithm for acyclic Temporal Bayesian network based on model simplification and general probability updating algorithm for Temporal Bayesian network are presented. While, study the Scheduling Bayesian network to model software project scheduling risk. The modeling method, related calculation and probability inference algorithm are presented.Temporal Bayesian network and Scheduling Bayesian network provide modeling method and inference algorithms for the management of software project iterative process risk.