改进的遗传算法在测试数据自动生成中的应用  

Improved adaptive GA application on automatic test data generation

作　　者：高海昌[1] 冯博琴[1] 朱利[2] 郭斌[1] 

机构地区：[1]西安交通大学电子与信息工程学院,陕西西安710049 [2]西安交通大学软件学院,陕西西安710049

出　　处：《系统工程与电子技术》2006年第7期1077-1081,共5页Systems Engineering and Electronics

基　　金：国家"863"高技术计划资助课题(2003AA1Z2610)

摘　　要：提出了图的一种新的存储方式:双向邻接链表,可完备地表示软件项目的类图结构。基于自动插装,构造了一种新颖的用树结构来设计适应度函数的改进遗传算法iAGA,并使用自适应的交叉和变异算子来进行下一代生成。将改进算法与基本遗传算法SGA和自适应遗传算法AGA进行了比较,发现可以明显节省搜索和生成时间。使用三角形分类等实例进行了测试数据自动生成的验证,证明所提的iAGA改进算法是有效的。对采用均匀权值、自顶向下和自底向上这三种权值分配方式进行了比较,实验表明在采用树结构计算适应度方式下,均匀分配权值比较。The current research of software test data automatic generation is not sufficiency. To express the structure of class chart perfectly, a novel storage strategy of chart, bidirectional adjacent list, is presented. An improved adaptive genetic algorithm （iAGA） using instrumentation technique based on information extraction from source file and a novel construct fitness function using tree structure are developed for automatic software test data generation. The performance between SGA, AGA and iAGA was compared. The experimental results show that the iAGA may significantly reduce the seeking time and generation and the iAGA is effective and promising. Average weight, Top-to-down, and Down-to-top assign methods are also compared, it shows that the average weight assign method is more excellent than others under tree structure fitness calculating.

关 键 词：测试数据 遗传算法 自适应算子 树结构 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]