Research On Intelligent Test Scenario Generation Based UML Activity Diagram

Software testing is the most important method for software quality and takes about 40%-50% of development cost. Quality and Generation Method of test case affect software reliability and cost. As UML and RUP have been widely used, Scenario test method based on UML activity diagram can be applied in early stage of software development. As a result defects of software can be found out in time. Generating Scenario from UML activity diagram has two ways. One is manual the other one is automatic. The problem of manual method is costly with low efficiency and reliability. The automatic approach is limited with all kinds of UML models in the result its application area is small.Aiming at the defects of automatic method for test case generation, we present the improved anti-ant-agent algorithm which can improve the resolving ability and increase the application in UML activity diagrams. The algorithm can deal with multinest fork join pair structure and loop structure in activity diagram. At first this paper analyzes the advantages and status of UML model in software testing, studies the concept and the expansion method of its measurability. Afterwards, the improved algorithm standardizes the designator of fork join pair, adds some data structures such as ant memory body, system loop body, system information body and so on, designs the method to generate test scenario for multinest loop structure, processes multinest fork join pair by recursion, and finally generate corresponding thin-thread tree.Finally, we design a complete implementation scheme of the improved algorithm via a concrete example which is realized and validated by C# programming language in the .NET environment. The paper presents an activity diagram example from software testing contract for Shanghai UMT line 6 (AFC), transforms it into XML file, and generates thin-thread tree from XML file by application of the improved anti-ant-agent algorithm. From the experiment results it can be generalized that the improved anti-ant-agent algorithm can parse more complex activity diagram and improve the test sufficiency.