基于支持向量机的测试用例生成方法研究

发布时间：2018-04-12 19:22

本文选题：软件测试 + 测试用例　；参考：《南昌航空大学》2017年硕士论文

【摘要】：在软件测试领域,如何有效地设计或生成符合要求的测试用例是一个十分重要的研究课题。在功能性测试的过程中,随着软件的功能需求不断增多,软件的规模逐渐变大,结构变得更加复杂,这可能会引发大量的隐藏故障。虽然目前有很多测试数据生成器可以自动生成大量的测试用例,但是这些测试用例都缺少了期望结果。如果采用人工的方式来为每个测试用例设计期望结果,不仅代价昂贵而且枯燥乏味,容易出错。论文介绍了软件测试的背景和基本概念,阐述了支持向量机(Support Vector Machine,SVM)的基本原理和相关知识;总结了功能性测试的方法,重点介绍了机器学习方法在软件测试中的应用;研究了自动生成输入数据集的测试工具(PICT)及其组合覆盖标准;构建了待测试软件的功能模型,并评估了其性能。本文在研究输入数据自动生成的基础上,依据待测试软件的输入输出关系构建了SVM模型,该模型可以用于预测测试用例的期望输出结果。因此,本文提出了一种包含期望输出结果的测试用例生成方法。该方法选取PICT测试工具产生输入参数的2-way组合或3-way组合作为典型样本集,并依据待测试软件的输入与输出关系,水平拼接出对应的输出结果。采用人工的方式,使得典型样本集覆盖了输出参数的所有组合,然后利用典型样本集来训练SVM模型。训练好的模型能够预测出n-way组合的输入数据对应的预期结果。最后,将该方法应用于“活动图的模型文件验证”实例中,采用Accuracy和Precision指标来评价预测的结果。实验结果表明,采用遗传算法进行参数优化,同时考虑输入域和输出域的参数组合覆盖,构建的功能模型具有更好的性能,从而说明了该方法的有效性。
[Abstract]:In the field of software testing, how to effectively design or generate test cases that meet the requirements is a very important research topic.In the process of functional testing, with the increasing functional requirements of software, the scale of software becomes larger and the structure becomes more complex, which may lead to a large number of hidden faults.Although there are many test data generators that can automatically generate a large number of test cases, these test cases lack the expected results.If the desired results are designed manually for each test case, they are expensive, boring and error-prone.This paper introduces the background and concept of software testing, expounds the basic principle and related knowledge of support vector machine support Vector machine, summarizes the methods of functional testing, and introduces the application of machine learning method in software testing.The test tool PICTand its combined coverage standard for automatically generating input data sets are studied, and the functional model of the software to be tested is constructed, and its performance is evaluated.On the basis of studying the automatic generation of input data, this paper constructs a SVM model based on the relation between input and output of the software to be tested, which can be used to predict the expected output results of test cases.Therefore, a test case generation method with expected output results is proposed.In this method, the 2-way combination or 3-way combination of input parameters produced by PICT testing tool is selected as the typical sample set, and the corresponding output results are stitched horizontally according to the relation between input and output of the software to be tested.The typical set of samples covers all the combinations of output parameters, and then the typical set of samples is used to train the SVM model.The trained model can predict the expected results corresponding to the input data of the n-way combination.The experimental results show that the proposed model has better performance by using genetic algorithm to optimize the parameters and considering the combined coverage of input and output fields, thus the effectiveness of the proposed method is demonstrated.
【学位授予单位】：南昌航空大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.53;TP18

【参考文献】