抄録
Measuring software product and the development process is essential for improving software productivity and quality. Software metrics are quantitative measures of software products and process. For example, complexity metrics are used to evaluate maintenance difficulty of a program. In order to evaluate the complexity of object-oriented program, several complexity metrics have been proposed. Among them, Chidamber and Kemerer's metrics are the most well-known metrics for object-oriented programs. Also, their effectiveness for estimating the number of faults in a program has been evaluated empirically. However, from the viewpoint of the management, it is more important to estimate the cost to fix the faults in the program rather than the number of ones. This paper empirically evaluates the usefulness of Chidamber's metrics by examining the relationship between the value of the metrics and the cost to fix the faults in the program by using the data collected from actual software development processes. This paper also examines how to apply the metrics to the programs most of which is developed by reusing the classes in the class-library (framework).
