Quantifying Skill and Contributor Risk in Software Projects: A Knowledge Unit-Based Analysis

This thesis examines the critical problem of quantifying the risk arising from the concentration of technical knowledge in a limited number of programmers in software development projects. In this context, an innovative methodology is proposed to identify and evaluate skills through source code analysis using machine learning techniques. The central concept of the study is the use of Knowledge Units (KUs), i.e. coherent sets of technological knowledge within the code, with application to Java projects. The paper extends existing KUs detection infrastructure and introduces two new risk metrics: the KU Risk index, which assesses the organizational risk resulting from the concentration of critical technologies in a limited number of individuals, and the Employee Risk index, which assesses the risk of losing expertise due to the potential departure of highly skilled developers. These metrics were applied to open source projects of the Apache and Eclipse organizations, leveraging historical data from commits and correlating KUs with their respective contributors. At the same time, an interactive visualisation tool was developed, which allows the monitoring of risk indicators at project, team or organisation level. The work provides an essential decision support tool for the software industry, contributing to the optimisation of human resource management, the early identification of technological dependencies and the strategic prevention of critical knowledge loss. Finally, it substantially strengthens the link between theoretical research and practical applications in the field of skills and know-how management.