A Comparative Study of Algorithms for Estimating Truck Factor

Abstract

In modern software projects, it is crucial to have reliable data about how knowledge on the source code is distributed among the team members. This information can help for example to avoid "islands of knowledge" and to prevent the risks associated to the loss of key developers. Truck factor is a key measure proposed to estimate such risks. Basically, truck factor (aka bus factor) designates the minimal number of developers that have to be hit by a truck (or quit) before a project is incapacitated. Although being a key measure of the concentration of information among team members, we still have few algorithms proposed to estimate truck factors. More importantly, we lack rigorous comparisons of the existing algorithms. Therefore, in this paper we provide a comparative study of the two main algorithms proposed in the literature to estimate truck factors. For this purpose, we rely on a large dataset of 133 popular GitHub systems. We compare both the performance of these algorithms and the truck factors estimated by them.