Comparing Machine Learning Correlations to Domain Experts’ Causal Knowledge: Employee Turnover Use Case

Meddeb, Eya, Bowers, Christopher ORCID: https://orcid.org/0000-0002-5076-512X and Nichol, Lynn (2022) Comparing Machine Learning Correlations to Domain Experts’ Causal Knowledge: Employee Turnover Use Case. In: Machine Learning and Knowledge Extraction, August 23–26, 2022, Vienna, Austria. ISSN Series ISSN 0302-9743; Online ISBN 978-3-031-14463-9

Text 978-3-031-14463-9_22 - Published Version Restricted to Repository staff only Download (152kB) | Request a copy

Abstract

This paper addresses two major phenomena, machine learning and causal knowledge discovery in the context of human resources management. First, we examine previous work analysing employee turnover predictions and the most important factors affecting these predictions using regular machine learning (ML) algorithms, we then interpret the results concluded from developing and testing different classification models using the IBM Human Resources (HR) data. Second, we explore an alternative process of extracting causal knowledge from semi-structured interviews with HR experts to form expert-derived causal graph (map). Through a comparison between the results concluded from using machine learning approaches and from interpreting findings of the interviews, we explore the benefits of adding domain experts’ causal knowledge to data knowledge. Recommendations are provided on the best methods and techniques to consider for causal graph learning to improve decision making.

Actions (login required)

View Item