基于决策树的软件项目估算方法  

A Software Project Estimation Method Based on Decision Trees

作　　者：阎巍[1] 

机构地区：[1]长沙学院计算机科学与技术系,湖南长沙410003

出　　处：《计算机工程与科学》2009年第8期117-120,共4页Computer Engineering & Science

摘　　要：软件工程领域的一个重要问题是预测软件开发项目的规模、工作量和成本,即软件项目估算问题。基于机器学习的方法在软件项目估算领域具有优势地位,本文提出了基于决策树的聚类分析预测方法,通过对目标项目的目标属性进行正确分类,预测目标属性的取值范围。通过对502个ISBSG v9项目数据集中的项目进行基于C4.5算法的分类预测,正确率达到82.4701%,满足了软件项目估算的指标要求。An important problem in the field of software engineering is to make predictions concerning the size effort/ cost of software development projects, namely software project estimation. The prediction methods based-on machine learning are more effective than other ones, and this paper presents an estimation method based-on decision trees, which is one of the commonly used machine learning tools. This technique predicts the value scope of the target project by classifying the ＇effort＇ attributes. We exam the use of decision trees in practice, and the results from a subset of the ISBSG v9 data set show that the demands of software project estimation are met.

关 键 词：软件项目估算 机器学习 决策树 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]