Explainable AI for Cost Estimation

  • Typ:Bachelor's thesis
  • Betreuer:

    Dominik Hammann

  • Zusatzfeld:

    2020

  • For the application of complex machine learning models, explanations
    and interpretability often play a crucial role besides accuracy. This is
    also the case in Cost Estimation, where the employment is often viewed
    sceptically due to the lacking credibility of models (Smith and Mason,
    1997, pp. 21-23). Thereby, questions arise like: How to capture a model's
    interpretability? How does interpretability cohere with accuracy? In which
    way contributes the research field of Explainable AI to those problems in
    general? This paper aims to answer these questions with an analysis of
    interpretability in the literature, a simulated use case for cost estimation
    and an empirical survey. Finally, an approach for a twofold evaluation
    of both model accuracy and interpretability for cost estimation models is
    proposed.