A Study On The Risk Decision Rules Mining Of IT Project Based On Rough Set

With the development of strategy of Informationization drives industrialization in China, the construction of informationization has become the main approach to enhance core competence for enterprises. Therefore, IT projects, as a part of informationization, have been paid more and more attention by people. However, implementation of IT project has been considered to be full of risk, and the rate of success is always very low. To a large degree, the success of project depends upon the valid identification, forecast and control of the risks. Hence, risk analysis and decision-making of IT project have been important problems in current project management. During the implementation of IT project, there are various risk factors with many characteristics such as uncertainty and structureless, so it is difficult to resolve all of risks with a single model and algorithm. However, traditional project risk analysis and decision-making are based on managers'experience and assumed many conditions, thus there are great differences between the result of analysis and actual situation. Thereby the traditional method has certain subjectivity and randomness. Aiming at these problems, this paper makes systematically and deeply researches on risks of IT project through questionnaire. Some new rules mining methods which are mainly based on Rough set theory and various approaches, are proposed in this paper, with the purpose that in the condition of incomplete and inaccurate information decision-maker can find concealed knowledge from in large amount of data directly without the traditional complicated process of building model. Thus it can be realized risk rank forecast and rules mining of risk decision during the IT project implementation, so that improve the ability of risk decision, promote risk management of IT project towards intelligent ,scientific and quantitative process, and increase the success rate of IT project finally. In this paper, the problems of rules mining during the risk analysis and risk decision of IT project are discuessed, and main parts are as follows: First of all, after a review of the current situation about IT project risk management in China and other countries, the scope and definition of IT project are described and analyzed, and according to the result of questionnaire, the main exhibitions of a failed IT project is summarized, and the risk factors in IT project are classified. Secondly, the data of risk factors can be pretreatment using method of Rough set, mainly including knowledge classification, classifying quality, and the mutual dependence among attributes. After pointing out the shortcoming of the single weight, it analyzes the combined weight of multi-attribute and clustering analysis of risk factors, with bringing the conception of combined weight, constructs similarity coefficient matrix Rn×n of combined weight using the distance formula of Euclid, so that it uncovers the relationship among risk factors,and realizes clustering analysis of risk factors. Thirdly, risk rule mining method based on Rough set standard classification is analyzed. First of all, the rule mining algorithm of classification consistency based on Rough set is introduced with examples (RICCR).based on RICCR, utilizing discernibility matrix CD(i, j) and discernibility function, it can simplify the process of attribute reduction and core values calculated. In the phase of simplifying rules, using the method of valid rules weight, it attains the goal of rule mining. As to decision system of classifying inconsistency, an improved method of rule mining algorithm based on classifying inconsistency is proposed, certainty and possibility rules are obtained by R*(x) and R*(x). Fourthly, it researches methods of decision rules mining of IT project risk in the condition of incomplete information. A similar null valuation model based on Rough set is constructed, and illustrates a rule mining method which is based on similarity relation of Rough set and explained by a real instance. Then, default rules in the process of risk decision can be mined using the method combined reduced lattice and Rough set. With the given μ, the rule in each node can be found out with this method. Furthermore, the decision rules can be optimized by priority principle. Fifthly, the dynamic risk decision rule mining is analyzed in IT project. shortcomings of the reduction method ofγcriterion based on Rough set are pointed out because of lack of considering the dynamic changes of risk facotrs. Therefore, this paper proposes modification ofγcriterion, namely information entropy criterion, which analyzes dynamic rule mining method based on information entropy from two aspects of Hdet (d /Q) and H loc (d /Q). At last, it demonstrates reasoning process of rule change and approximate dynamic rules, so it overcomes the shortcomings which exist in the process of obtaining static rules, and avoids the disadvantages of information distortion and rule change due to noise pollution.Finally, the rule mining method combined the Rough set and Bayesian theorem is discussed deeply in this section. There are shortcomings only using Bayesian theorem or Rough set to deal with incomplete information, then analyzes Bayesian classifiers in detail, combines Rough set method and NB Bayesian classifiers, makes data reduction utilizing Rough set firstly, and then reduction the data can be trained by NB Bayesian classifiers secondly. Thus, the Rough set-Bayesian classifiers not only reduce the size of data, but also obtain the ability of classifying incomplete data and learning new knowledge with increment. So on one hand, it avoids the disadvantage of the Bayesian classifiers being dependant upon experience overly, on the other hand, it overcomes the shortcoming of using Rough set completely to deduce rules. It indicates that integrative rule mining system that is constructed by combining different methods can overcome individual's limitation and the function is more powerful than the single system.