协同过率在工作量估算中的应用

摘要

工作量估算是软件项目管理的重要内容之一，协同过滤是一种在历史数据中确定相似用户或物品产生推荐的方法，已成功的应用于电子商务、影视推荐等多个领域，本文将协同过滤技术应用于软件工作量的估算。首先对历史项目集中的数值型数据进行归一化，然后采用均值对缺失值进行插补，余弦用于计算项目的相似度，最后确定项目的近邻集对待评估项目的工作量进行估算。从USP05-FT中选择了4个项目作为实例来说明该估算过程，估算结果与实际值有一些偏差是由于协同过滤仅能处理数值型数据。%Software effort estimation is an important aspect of project management, collaborative filtering is a method that produces recommend based on similarity of user or items in the historical data It has been successfully applied to e-commerce, video recommendations, and other ifelds, collaborative ifltering technology is applied to software cost estimation in the paper. Firstly, the data of different types is normalized;then missing values are inputted by average value and similarity of items is computed using cosine;lastly;the project’s neighbors are found to estimate effort. 4 projects data from USP05-FT are selected as an example to describe the estimation process. The deviation between estimates and the actual value derive from it that collaborative ifltering can only handle numerical data.