Absenteeism Prediction: A Comparative Study Using Machine Learning Models

Abstract

Solidity of companies or institutions is related to several factors but mostly to absenteeism. Taking annual leave or pre-determined absent days of personnel may be covered by others however, unexpected absenteeism causes irredeemably poor results. Prediction of the correlation between this predetermined and unexpected absenteeism is a challenging task and includes nonlinear relationship. Neural Network based Machine Learning models are built to solve this kind of non-linear problems by using their non-deterministic nature. In this research, three neural network models; Backpropagation, Radial Basis Function and Long-Short Term Memory neural networks, are implemented to solve prediction problem of absenteeism. In addition, a comparative study is conducted between these models. Two experiments with different training ratios and three evaluation criteria are considered and implemented. The experimental results suggested that Long-Short Term Memory neural network has very high prediction rates as 99.9% in prediction problems that consists complex data and it produced superior results than other two neural network models.