Uncovering Key Drivers of Absenteeism in Manufacturing: A Machine Learning-Based Feature Selection Study

Abstract

Absenteeism management is a critical issue for manufacturing companies that rely heavily on direct human labor for production. High absenteeism rates can significantly hinder operational productivity. This paper aims to identify the most influential features affecting absenteeism in a textile production plant. The study begins with thorough data preprocessing, including handling missing values and outliers. Subsequently, various analytical techniques, particularly from the field of machine learning, are applied to determine the most relevant predictors of absenteeism. Specifically, we apply four complementary feature selection techniques: Mutual Information, Correlation Analysis, Recursive Feature Elimination (RFE), and Particle Swarm Optimization (PSO). The results reveal that 9 out of 36 features play a significant role in explaining absenteeism patterns. Identifying these key variables provides valuable insights for decision-makers to better target their efforts and enhance absenteeism management strategies.