基于改进花朵授粉算法的测试数据自动生成研究

发布时间：2018-10-30 12:05

【摘要】：测试数据作为软件测试中的核心因素,其生成效率高低直接影响着软件测试的效果,本文主要对测试数据的自动生成方法进行研究。花朵授粉算法作为一种具有良好寻优能力的新型智能算法,因其参数简单、容易实现,已被成功应用于各种多目标优化问题中,本文研究将此算法应用于测试数据自动生成领域,首先针对其缺陷提出一系列改进措施,再通过大量实验验证改进后的算法在测试数据自动生成中的优越性。本文工作内容与创新点主要包括以下几个方面:(1)针对基本花朵授粉算法搜索速度较慢、寻优精度不高和在中后期容易陷入局部极值的缺陷,从对算法参数进行调整和引入其他智能算法进行混合两大方向对基本花朵授粉算法加以改进,提出一种自适应混合花朵授粉算法,首先引入粒子群算法,利用粒子群算法在搜索初期阶段收敛精度高与速度快的优势获得一组质量较优的解作为花朵授粉算法的初始解来继续实施迭代寻优操作;其次,提出一个警卫函数来对反映种群的离散程度;最后采取一种自适应机制对解更新,自适应机制包括自适应柯西变异与自适应步长因子两部分,根据当前种群的离散程度大小以及解的位置状态自适应地进行寻优搜索,从而提高寻优能力。(2)对基本花朵授粉算法应用于测试数据生成上的理论依据进行分析,在此基础上研究如何将本文提出的自适应混合花朵授粉算法应用于测试数据的自动生成中,建立基于自适应混合花朵授粉算法的测试数据生成模型,同时提出了一种改进的适应度函数构造方法,通过分支被覆盖的难易程度不同来对每条分支分配相对应的权重参数值,以更准确地反映分支的覆盖情况,从而进一步提高测试数据的生成效率。(3)最后对本文提出的自适应混合花朵授粉算法在测试数据自动生成中的可行性与效率性进行验证,选取4种复杂程度不同的典型测试程序,借助MATLAB平台对其实现测试数据自动生成,同已经应用于测试数据自动生成领域的另外两种智能算法相比较,从平均耗费时间、平均迭代次数和平均分支覆盖比例3项数据指标进行对比分析。
[Abstract]:As the core factor in software testing, test data generation efficiency directly affects the effect of software testing. In this paper, the automatic generation method of test data is studied. As a new intelligent algorithm with good optimization ability, flower pollination algorithm has been successfully applied to various multi-objective optimization problems because of its simple parameters and easy implementation. This paper studies the application of this algorithm in the field of automatic generation of test data. Firstly, a series of improvement measures are put forward in view of its defects, and then the superiority of the improved algorithm in automatic generation of test data is verified by a large number of experiments. The contents and innovations of this paper mainly include the following aspects: (1) aiming at the defects of the basic flower pollination algorithm, the search speed is slow, the searching accuracy is not high and the local extremum is easy to fall into in the middle and late period. An adaptive hybrid flower pollination algorithm is proposed by improving the basic flower pollination algorithm by adjusting the parameters of the algorithm and introducing other intelligent algorithms into the hybrid algorithm. Firstly, the particle swarm optimization algorithm is introduced. Based on the advantages of high convergence accuracy and high speed in the initial stage of searching, a group of better quality solutions are obtained as the initial solution of the flower pollination algorithm to continue the iterative optimization operation. Secondly, a security function is proposed to reflect the discrete degree of the population. Finally, an adaptive mechanism is adopted to update the solution. The adaptive mechanism consists of two parts: adaptive Cauchy mutation and adaptive step size factor. According to the size of the population dispersion and the location of the solution, the adaptive search is carried out adaptively. In order to improve the ability of optimization. (2) the theoretical basis of basic flower pollination algorithm applied to test data generation is analyzed. On this basis, we study how to apply the adaptive hybrid flower pollination algorithm to the automatic generation of test data, and establish a test data generation model based on adaptive hybrid flower pollination algorithm. At the same time, an improved fitness function construction method is proposed to assign the corresponding weight parameter values to each branch according to the difficulty and ease of each branch, so as to reflect the coverage of each branch more accurately. In order to further improve the efficiency of test data generation. (3) finally, the feasibility and efficiency of adaptive hybrid flower pollination algorithm in automatic generation of test data are verified. Four typical test programs with different degrees of complexity are selected to generate test data automatically with the help of MATLAB platform. Compared with the other two intelligent algorithms that have been applied in the field of automatic generation of test data, the average time consumption is obtained. The average iteration times and the average branch coverage ratio were compared and analyzed.
【学位授予单位】：江西理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.53;TP18

【参考文献】