The Research Of Java Software Clustering Based On Extended Dependency Model

Software engineering practice shows that, if the structure of software can match the design feature of “high cohesion, low coupling”, the software can not only due to the ripple effect caused by a change of control in the smaller range, but also reduce the cost of software maintenance later. However, the design of software is a complex process, software developers often lack a deep understanding about the problem areas, so the early design may not fully meet the design feature of “high cohesion, low coupling”. In addition, it may appear inconsistent between the code implementation and design documentation, so need way to get the architecture from source code and verify that whether the architecture is same with the design feature.Research objective of this paper is that reverse exporting structure from Java source code, and verifying whether the software structure compliance with the design feature of “high cohesion, low coupling”. The main contents include:1, To obtain the co-change relationship among Java classes from multiple versions of source code, and get the attribute information from the co-change relationship in software evolution process;2, Building extended dependency model using the attribute information.3, Software Clustering based on the extended dependency model;4,Designing and implementing a system based on the above.Based on the above objectives and content, this paper presents that the Research of Java software clustering based on extended dependency model. Firstly, it describes that the extended dependency model and how to build the extended dependency model. The extended model is adding the attribute information in software evolution process in the original model, and we should get the attribute information from the co-change relationship in software evolution process. The steps for building the extended dependency model include: Getting real property set, Extracting co-change relationship among Java classes, Calculating the similarity between entities and so on. Secondly, it describes the software clustering based on the extended dependency model in detail, and it describes the experimental procedure about two cases of no-extended dependency model and extended dependency model. Finally, it describes the design and implement of the system. The innovation of this paper include:1、Exracted Java class co-change by association rule algorithm,and builded the extended dependency model on the basis;2 、 Software clustering based on the extended dependency model;3、Designed and implemented a software system for clustering based on the above model.