Using the AVL Algorithm to Evaluate the Bus Factor of Research Software Projects
Ellie Kobak
Committee: Boyana Norris
Honors Bachelors Thesis(Sep 2022)
Keywords:

This paper evaluates the accuracy of AVL Bus Factor (BF) estimations using commits through comparing the results to lines of code changes (LOCC) and the cosine difference of lines of code. In order to compare the BF values, the AVL BF algorithm for all three metrics was run on five open-sourced high performance computing projects. Three of the projects were then evaluated over a five-year span to determine any trends of BF across projects. The results across all five projects showed different values for the BF when using LOCC and cosine difference, with LOCC typically having a higher BF than commits and cosine difference. When comparing the three metrics on a year-to-year basis, the commits BF value was typically higher than the other two metrics. The implications of the commits having a higher BF estimation can lead to false sense of stability within projects, leading to the potential of more issues arising.