代码挖掘中的数据处理方法综述

Survey on Data Processing in Code Mining

程序代码中蕴含着软件开发人员最原始的开发理念、设计思想和编程习惯等信息,将数据挖掘用于分析处理这种编码痕迹以便提取出潜藏着的有用知识是一个有着广阔前景的新的研究领域.由于当前的挖掘程序尚无法直接处理这种文本结构的程序代码,因而研究者需要将软件代码抽象成一种更有效的中间表达形式来作为挖掘对象.这种中间表达形式不仅界定了挖掘所使用的算法,更重要的是,它决定了所能挖掘的知识内容.对代码挖掘的一般过程进行了介绍并着重分析了代码挖掘的各种中间表达形式的特点.在此基础上,指出了当前代码挖掘存在的问题及今后的发展方向.

Source code contains a wide range of information such as the original concepts, design ideas and programming habits of the developers. Analyzing coding traces in source code with data mining technology to extract those hidden but useful knowledge has been a new but promising research field. Since the current mining applications could not handle with source code of text structures, it is usually necessary to abstract software code into a more effective form of expression called intermediate form to simplify the mining process. Such expressing forms decide not only the mining algorithm used but also, more importantly, decide which kind of knowledge could be extracted from. This paper anatomized the general process of data mining based on source code and focused on the analysis of specialty of kinds of intermediate forms. On this foundation, it put forward the current problem in this area and point out the future direction of development.