Statistical Software Testing Based On Markov Chain Usage Models

Statistical software testing based on a Markov chain model builds at first the Markov chain usage model which describes usage profiles of software and then generates test cases from the usage model, applies them to the software, finally estimates the reliability and quality of the software according to statistical models. The Markov chain usage model presents the profile of actual or anticipated use of software with a stochastic model. Some model statistics analysis can aid in test planning and model verification. The usage model allows software testers to generate automatically or manually test sequences using some statistical testing tools. The problem is how to construct the Markov chain usage model for the statistical testing. In practice, the assignment, verification and optimization of transition probabilities of the usage model can be very difficult. In this thesis, we analyze various methods existing for establishing automatically transition probabilities of usage models, and propose new methods. The content of this thesis could be included in these areas:1. Have an overall summarization the of research actuality. Introduce the main methods of software testing. Discuss the difficulties of the construction of the Markov chain usage model in statistical software testing.2. Analyze the importance of Markov chain usage model for the test planning. Discuss the construction and representation of the Markov chain usage model. Analyze the statistics of usage model and its senses for test planning, study the generation of test cases from the Markov chain usage model.3. Analyze various methods existing for establishing and optimization automatically transition probabilities of Markov chain usage models. Improve the method of S. A. Sherer. The improved method uses the values of failure loss, generates automatically transition probabilities subject to transition probabilities constrains by decreasing failure loss. Propose a new method to generate automatically transition probabilities of usage models subject to transition probabilities constrains by minimizing the complexity of the Markov chain usage model.4. Develop a statistical testing tool based on C. It supports model files of format TML reading, statistics analysis, test cases generation. It also supports generations of transition probabilities of usage models subject to transition probabilities constrains by decreasing failure loss or by minimizing the complexity of the Markov chain usage model.