软件相似度在成本估算中的应用

摘要

Accurate cost prediction is one of the important contents in project management.The analogy-based estimation is an important method in project cost estimation,the accuracy of estimation has direct relation to the selection of similar project.Because of this,we studied the similarity of software projects.We applied the idea of sentence similarity to the calculation of software similarity,first processed the missing value in dataset,and then calculated the similarities between each property of the project,adopted different method in calculation for different property valuing types,and finally obtained the similarities between the projects by making weighting mean on the similarities between each property.We also applied this method on USP05-FT project set for cost estimation.Experimental result showed that the average accuracy of estimation can reach over 85% by selecting proper similarity for the selection of near neighbours.%准确的成本估算是项目管理的重要内容之一，基于类比的估算是项目成本估算的一种重要方法，估算的准确性与相似项目的选择有着直接的关系。鉴于此，对软件项目的相似度进行研究。将语句相似度的思想用于软件相似度的计算，首先对数据集中的缺失值进行处理，然后计算项目各属性之间的相似度，针对不同的属性取值类型采用不同的方法进行计算，最后将各属性间的相似度进行加权平均得出项目间的相似度。将该方法应用于 USP05-FT 项目集进行成本的估算。实验结果表明：通过选择合适的相似度进行近邻的选择，评估的平均准确性可以达到85％以上。