A novel strategy for automatic test data generation using soft computing technique

Software testing is one of the most crucial and analytical aspect to assure that developed software meets pre- scribed quality standards. Software development process in- vests at least 50% of the total cost in software testing process. Optimum and efficacious test data design of software is an important and challenging activity due to the nonlinear struc- ture of software. Moreover, test case type and scope deter- mines the quality of test data. To address this issue, software testing tools should employ intelligence based soft comput- ing techniques like particle swarm optimization （PSO） and genetic algorithm （GA） to generate smart and efficient test data automatically. This paper presents a hybrid PSO and GA based heuristic for automatic generation of test suites. In this paper, we described the design and implementation of the proposed strategy and evaluated our model by performing ex- periments with ten container classes from the Java standard library. We analyzed our algorithm statistically with test ad- equacy criterion as branch coverage. The performance ade- quacy criterion is taken as percentage coverage per unit time and percentage of faults detected by the generated test data. We have compared our work with the heuristic based upon GA, PSO, existing hybrid strategies based on GA and PSO and memetic algorithm. The results showed that the test case generation is efficient in our work.

基于改进粒子群算法的测试数据自动生成研究

测试数据的自动产生技术是软件测试的一个重要研究领域,高效的测试数据生成可以简化测试工作提高测试效率;针对传统遗传算法(GA)容易产生早熟收敛和易陷入局部最优解的问题,提出了一种基于遗传粒子群混合算法(GA-PSO)的软件测试数据自动生成算法,该算法在粒子群算法的基础上引入了遗传算子(交叉概率Pc、变异概率Pm),使所有测试数据在局部区域中再次寻找最优值,从而避免过早收敛,改进搜索最佳值的能力;仿真实验表明遗传粒子群混合算法与单纯使用简单遗传算法相比,具有更快的收敛速度,其产生最优解的代数得以大大提前,且精度更高。

基于粒子群优化算法的软件测试数据生成研究

遗传算法在软件测试数据自动生成方面应用广泛,但是其自身也存在局限性,如参数难于设置、算法复杂等,而粒子群优化算法执行容易、参数少,能很快地找到最优解。论文提出一种基于粒子群优化算法的软件测试数据自动生成方法,并应用于等边三角形判别程序。实验表明,粒子群优化算法能比遗传算法更高效的生成测试数据。

一种面向对象测试用例自动生成的混合算法

遗传算法(GA)在测试用例生成方面是一种实用的算法,但是其自身也存在的局限性,如过早收敛、优化效率低等问题。通过引入粒子群算法(PSO),使每一个测试用例在局部区域中再次寻找最优值,以此改进整体算法搜索最佳值的能力,避免过早收敛、优化效率低的问题。与此同时,针对面向对象测试的特点,如封装性等,将混合算法进行适当的改进,满足在不同环境中重复使用类的要求。