Researches Of Software Requirements Modeling And Prioritization Based On Dependencies

In the face of fierce competition in the software market,the accuracy of requirements modeling and prioritization are key factors which affect the success of the software product.Requirements modeling and prioritization are key information to support software designing.But in current requirements modeling and prioritization methods,there are still some challenges to solve:1)ignoring the requirements dependencies;2)ignoring the modeling of non-functional requirements;3)the lacking utilization of domain knowledge;4)the accuracy of the prioritization depends too much on personal experience and subjective cognition;5)when the number of requirements increases,the operations of modeling and prioritization are complicated and inefficient.In order to handle above challenges,my research contents are as follows:(1)A method for constructing domain knowledge based on non-functional attributes of software is proposed,which is called NFADKC(Non-Functional Attributes based Domain Knowledge Construction).In this method,a kind of domain oriented software non-functional attributes relation model NFADKM(Non-Functional Attributes based Domain Knowledge Model)can be built.First of all,the international standard IS025010:2011 and CC(Common Criteria)are integrated to build the software non-functional attributes decomposition tree NFAT(Non-Functional Attributes Tree),which is used to help the classification of non-functional requirements;secondly,on the basis of the triangular fuzzy number method,a qualitative and quantitative data conversion way is presented to support quantification of the evaluation value,thus supportting the domain knowledge polymerization and domain knowledge analysis;then,a transfer function is defined based on the matrix of intersection,which can be used to obtain new requirements dependencies,and realize the expansion of domain knowledge.(2)A non-functional requirements modeling method based on NFADKM is proposed.On the basis of the NFADKM,some constraint rules are defined for helping the requirements engineers discovering more potential dependencies in requirements model,and improving the operability of the NFR framework.The crossover experiment shows that this method can enhance the efficiency of the traditional NFR framework,and improve the accuracy of requirements' modeling.(3)A novel requirements prioritization method based on requirements dependencies and user preference is proposed,which is called DRank.According to the evaluation attributes in current prioritization methods,we carry out the collation and classification of the evaluation attributes,and finally a unified evaluation attributes model PEAT(Prioritization Evaluation Attributes Tree)is puted forward.The machine learning algorithm RankBoost is used to calculate and integrate the users' preferences.The link analysis algorithm PageRank is used to determine the requirements'influences;finally,by combining users' preferences and requirements' influences with the business logical relationship between requirements,the synthesized requirements prioritization can be generated.(4)A controlled experiment is conducted for comparing DRank with three other methods,CBRank,AHP and EVOLVE.In the experiment,in order to evaluate the efficiency and accuracy of these four methods,the related evaluation parameters are extended.From the perspective of perception,related questionnaire surveies about“ease of learning”,“ease of use”and“usefuleness”are designed.At the same time,from the perspective of avaibility,the time-consuming,top matching rate and inverse rate for each method are recorded.The results show that the DRank is generally considered useful and easier to use,and the DRank is superior to CBRank,AHP and EVOLVE in terms of the accuracy of the prioritization.(5)A prototype system is developed to support requirements modeling and prioritization.The client provides convenience for users to submit individual knowledge,model requirements and prioritise requirements.The architecture and functions of the system are described in detail.