Using Network Analysis for Recommendation of Central Software Classes

Program comprehension is a complex task, especially for large software systems. Understanding an unknown system requires a significant amount of time. To speed up the learning process, developers focus on understanding central classes first. If other developers are available, they usually suggest which classes should be read at the beginning. In the absence of this knowledge, an independent algorithm is needed to measure the centrality of a class and give a recommendation for the developer. This paper presents an approach to retrieve central classes by using network analysis on the dependency graph of the system. An empirical study on open source projects evaluates the results of our algorithm based on a survey among the system’s developers.      «

Program comprehension is a complex task, especially for large software systems. Understanding an unknown system requires a significant amount of time. To speed up the learning process, developers focus on understanding central classes first. If other developers are available, they usually suggest which classes should be read at the beginning. In the absence of this knowledge, an independent algorithm is needed to measure the centrality of a class and give a recommendation for the develope...     »