Customer Requirements Analysis in Machine Learning Projects

Abstract

This Bachelor Thesis explores the various requirements of Machine Learning projects that utilize numerical data in their models. By interviewing various experts in the fields; varying in expertise, age, gender and ethnicity; we discover the key factors that often represent the biggest obstacles for the teams working with such technologies. The interviews were conducted between September and December of 2019 and studied until February of 2020. Using 2 research strategies: Interviews and Literature Reviews from news media articles and published academic work. The analysis was then executed through the recording of the interviews, their transcription and analysis conducted in tandem with the review of literature focused on Machine learning models and numerical data.

This thesis gives a general understanding into the work that is typically required these days to run these Machine Learning models. Uncovering that the problems more often originates from the lack of understanding from those wanting these technologies to be implemented in their systems that from the systems themselves. Highlighting the importance of understanding data and good practices when it comes to storing, structuring and modelling for ML models.