Research On Location Of Software Neglected Condition Defects Based On Maximal Sub-graph Mining Algorithm

In recent years, with the rapid development of application requirements andinformation technology, social production and life is more and more inseparable from allkinds of powerful software. There are a large number of defects in the software withcomplex structure and a large amount of code, and the neglected condition is the mainform of those defects. Moreover, the probability of false positives and false negative islarge and the efficiency is relatively low when using the existing test tools to detect defects.Therefore, the use of implicit programming rules to detect and locate the neglectedcondition defects in the software has become an important research direction in softwaretesting, because of its short time testing, high efficiency and high automation degree.The software defect localization system based on maximal frequent sub-graphslocates the neglected condition defects in the software through three steps, includingconstructing program dependency graph, mining the implicit programming rules andmatching the source code with rules. Specifically, the source code which is used fortraining will be unified to simplify the complex statements in the process of programdependency graph generation. And program dependency graph will be constructed byobtaining the data dependencies and control dependencies between the statements in theunified code, so as to minimize interference with the different style of programming. Thenthe vertices of program dependency graph will be labeled by using statement abstraction,and the program dependency graph for later mining will be formed at the same time.Moreover, the graph of implicit programming rules will be formatted by constructingmaximal frequent sub-graphs through using efficient graph mining algorithm to excavatethe denied program dependency graph and filtering maximal frequent sub-graphs throughusing necessary conditions of rule formation. Finally, the defect location will be realized byfollowing steps, including matching the program dependency graph with the graph of implicit programming rules seriatim to find the maximum similarity graph, and comparingdifferences of control dependency between the program dependency graph and the graphof optimal rule.The experiments show that the software defect localization method based on maximalfrequent sub-graph mining has better efficiency and effect, and can precisely locate thestatement of the source code where software defect takes place.