부산대학교 도서관

Independent Testing of business applications in the enterprise is largely a manual exercise. Automation, if any, is observed in test management and to a lesser degree in test automation. Test design comprising test architecture, test strategy, test procedure and test data is largely a manual activity. It is a common practice to express test cases manually as test scripts that lay down the test procedure in terms of instructions to testers. As systems evolve, test scripts are modified and if need be, new test scripts written. During maintenance, comprehension of test scripts for the underlying architecture and test strategies is important to affect changes. The problem is acute for IT service providers, providing independent testing services, as test scripts are inherited from client organizations and original authors may no longer be available. In this paper we propose a novel idea of mining process models and behavioral architectural components from test scripts. We have carried out preliminary investigations into mining process models from test scripts and present a set of challenges that need further investigations. We also present a scheme to classify test steps as a first step to mine architectural components. Experimental results of automatic classification using supervised learning techniques indicate accuracy between 88% to 94% motivating us to carry out further investigations.