The Construction Of Fuzzy SPRE Model And The Platform Development Based On Bayesian Network

The risk management of software project is an indispensable part in the cycle ofsoftware life, it identifies the potential risks as early as possible, adopts correspondingpreventive measures and makes plans to eliminate underlying risks or reduce theimpact brought by risk events. As the key factors of risk management, the riskanalysis and assessment have a great influence on the risk control, the cycle ofsoftware development and the product quantity. In recent years, software risk analysismethods are fastened on single-level qualitative analysis, and the relativity of risks isusually ignored. However, the software risk assessment is depended upon humansince no enough experience data is available, and meanwhile, there exists problemssuch as language fuzzy, which make it more difficult to measure risk parameters withtraditional statistics. Thus, traditional risk analysis methods and assessmenttechniques are far away from request for risk dynamic alteration simulation in realsoftware process.In 1982, Pearl began the research on probability reasoning with Bayesiannetwork in the field of artificial intelligence. Bayesian network not only expresses thedependences of variables well, but also deals greatly with the uncertainty brought bythe correlation among different knowledge elements based on theory of probability,and provides a method of diagrammatizing knowledge intuitively. Hence, Bayesiannetwork is more and more applied to the fields of expert system, data mining anddecision support. In the practical development and management of software project,there also exist relativities and causalities among risk affairs, risk probability, as animportant attribute of risk assessment, whose reasoning flat establishes on the riskdependence analysis, so modeling the risk analysis process with Bayesian networksatisfies the requirement perfectly.In 1965, Zadeh, an expert of cybernetics in America, put forward the fuzzytheory to describe the fuzzy phenomena in real world. Because of the abound fuzzyphenomena, the fuzzy theory was broadly applied in math and other utility fieldsaccomplished with its birth, especially made great contributions to the developmentof intelligent computer. As a result of the fact that the assessment of software riskdepends on human to great extent, and the language itself also has kind of fuzzy,more difficulties have been brought in the quantitative evaluation for experts. And thefuzzy theory adopts the representation of membership functions, and uses themulti-value logic to complete the operation and approximation of fuzzy numbers,which solves the uncertainty in risk assessment.On the basis of analyzing and researching the domestic and international currentsituation and existing problem, this paper presents a novel thought to establish thesoftware project risk evaluation model, the aim is to introduce concepts of the fuzzytheory, risk weight, and so on. In the process of the software development, the modelmeasures the possibility and the damage of risks objectively and effectively, andcalculates the combinative influence of various consequences and the disaster to thewhole result brought by risk factors, afterwards puts forward corresponding riskpreventive measure to reduce the damage of software risks and improve productquality as possible as we can. At the same time, combined with the characteristics ofsoftware enterprises in China, aiming at the long-term persistent development, thispaper adds the risk study mechanism to the model, and customizes the risk casedatabase according to actual circumstances for the enterprises.According to the above theory, this paper establishes the analysis and fuzzyevaluation model for the software project risks based on Bayesian theory, combinedwith initial risk classification system, and integrated with the independent developedplotting system, implements an risk analysis and evaluation expert system forsoftware project. It also introduces the framework, function flow and the databasedesign of the system and the methods to realize the key techniques of the systemincluding the data structure of risk analysis network model, the detect arithmetic ofrisk network structure, the dynamic storage of risk condition probability table, theprobability reasoning arithmetic of risk Bayesian analysis network, and the study ofrisk case structure. In the end, the feasibility and advantages of the software isvalidated with an instance of development course for software project.With the rapid development of software, the scale of the software products islarger and larger, and the problems existed in the software development process isincreased day by day, therefore, how to improve the quality of software products hasbecome the focus of the software field, which is urgent to be solved, and the existenceof software risk directly affects the quality of software products, or even the failure ofsoftware projects. Thus, the development of a risk evaluation system for softwareproject, which is suitable for the conditions of our country, is significant, it willdefinitely accelerate the software industrialization of our country.