Change Impact Analysis Using Linked Data Of Software Engineering

Change is the primary law of system engineering.During software development and maintenance,software entities are often changed because of the variation of requirements,environments and resources,or solution improvement.A change of one entity(such as one requirement,one class,and one method)often directly or indirectly impacts on other entities,and there will be a serious of changes,which is called the ripple efforts.Researchers have proposed several methods to analyze software change impact,to identify the potential impact of partial change on other parts of the whole system.Most of previous impact analysis technologies only focuse on single software artifact,such as source code or requirement models.But all the software artifacts compose of a whole organic.So the software change impact analysis should cover the whole software life cycle.Under this situation,this paper proposes a new approach of change impact analysis using linked data of software engineering.It analyzes internal or external associations of requirements,classes,program files,bug reports,code commit history and developers to build linked data of software engineering.It researches on change impact analysis for multi-artifacts,and effectively solves data problems,such as the complicated description way,information inconsistency and data imperfection.It extracts relational features from these linked data,to build feature model with weights.Finally,it calculates the change impact propagation degree to get change impact set with higher accuracy.The main contributions of this paper are as follows:1)Refinement analysis of software engineering linked data.It analyzes the association of multi-artifacts,to get software engineering linked data.It analyzes class similarity,bug reports similarity and direct relationships between requirements deeply.2)Relational feature modeling of change impact analysis.It extracts direct and indirect relational features from software engineering linked data.Considering different impacts by different associations,it sets one weight for each relational feature,to build relational feature model.Then an experiment is carried to calculate weights with the help of WEKA using logistic regression.3)Chang impact propagation calculation based on random walk.The impact degree between entities can be calculated by relational feature model,which can formulate degree matrix.The algorithm based on random walk handles the matrix and get different change impact set with different change position.4)Experiments.Several change impact analysis experiments are made using data from two open source projects,HtmlUnit and OpenRocket.The experimental results show that compared with other methods,the proposed method can get better precision,recall and F-measure.It effectively solves the problems such as information tracing of artifacts with different levels and low accuracy of change impact analysis for multi-artifacts.