Learning from software reuse experience

摘要

What makes certain software reuse practices successful is not currently well understood. This is partly because experience with reuse is scarce. Furthermore, data on reuse experience often includes many qualitative factors. These factors limit the applicability of statistical analysis. This paper presents an empirical study that investigates and evaluates an approach to extracting information from past software reuse experience. Specifically, a machine learning system, SORCER is introduced and applied to survey data on 24 software reuse projects. The data has 29 attributes, and was obtained from 19 companies over a three-year period. Results show that, as expected, some management actions are crucial for success of a project. Interestingly, most factors influencing the implementation of reuse programs and company profiles relevant to reuse projects have little impact on project success. Instead, some of the reuse program implementation factors (e.g., reuse approach, independence of the reuse development and presence of domain analysis) appear to have impact on project failure.