Research On Requirements Trace Links Generation Method Based On Hybrid Information Retrieval Model

Requirements Trace links between requirements and software artifacts are important bases for requirements change impact analysis and requirements change decision-making.Therefore,it becomes a key factor that affects the success of development projects.Unfortunately,establishing requirements trace links is a tedious,labor-intensive,and fallible task.To alleviate this problem,Information Retrieval(IR)methods such as Vector Space Model(VSM),Latent Semantic Indexing(LSI),and their variants,have been widely used to establish trace links automatically.But with the widespread use of agile development methodology,artifacts that can be used to generate automatic tracing links are getting shorter and shorter,which decreases the effects of traditional automated requirements trace link generation methods.In this dissertation,we firstly analyze the advantages and disadvantages of the existing automated requirements trace links generation technologies.For the problem that artifacts are getting shorter and less in agile development methodology,we proposed a requirements traceability(RT)challenge-oriented paper recommendation method for emerging IR technology.Then,based on this technique,we proposed a hybrid IR-based method to generate requirements trace links.Specifically,this dissertation includes the following aspects:1)Propose a RT challenge-oriented paper recommendation method for emerging IR technology.This recommendation method uses the information in the master-slave(RT-IR)knowledge graph to constitute the characteristics of new technology papers that users want to retrieve,including the features of object challenges,object technologies and exclude technologies.Based on this knowledge,the technology papers with highest feature similarity will be recommended from the papers retrieved by simple keywords.Experiments show that the effect of the recommendation method has reached an acceptable level,which greatly improves the efficiency of finding new technology papers.This method provides a research paradigm for researchers to find challenge-oriented new technologies in other disciplines.2)Propose a hybrid IR-based requirements trace links generation method.It uses the adaptability to short-text artifacts of the biterm topic model and its complementary ability to the vector space model.After combining these two information retrieval models,a new method for generating requirements trace links is formed.This method not only improves adaptability for short-text artifacts,but also improves the effect baselines of IR-based tracing methods.Gratifyingly,it is the first time that BTM has been used to generate requirements trace links.3)Propose a fully automated and improved hybrid IR-based tracing method.This method not only improves adaptability for short-text artifacts and effect baselines of IR-based tracing methods,but also reduces the dependency on requirements engineer experience by introducing genetic algorithm to determine the initial parameter values for BTM.The experiments results show that it is a fully automated,trustworthy,and short text artifacts-support trace links generation method.It has the potential to transfer technology to the industry.