Predicting Unplanned employee Absenteeism

Abstract

Unplanned employee absenteeism presents a critical operational and financial challenge to organizations across industries, with 
U.S. employers losing approximately $225.8 billion annually. This paper proposes a specialized cloud-native machine learning 
framework to forecast shift-level absenteeism 12-48 hours in advance. Our proposed approach would integrate diverse data sources 
including historical attendance patterns, shift metadata, weather conditions, traffic data, and holiday proximity to create a robust 
predictive model. This proposal details the significant business problem, specific tree-based modelling decisions, hyper parameter 
optimization strategies, and implementation considerations that could enable substantial performance gains in enterprise workforce 
management.