The Application of Machine Learning in Predicting Absenteeism at Work

Abstract

The employee attendance is an important indicator to judge employees' work attitude and measure their workload, which directly impact the development of the corporation. Meanwhile, the absenteeism, referring to the employees' intentional or habitual absence from work, is also significant for the whole company. Having a good knowledge of the reasons and the predictions of employees' absenteeism can help the leaders to adjust the ways of working and be prepared to avoid the major effect on company finances, morale and other factors, which made by decreased productivity. While the leaders may judge the absenteeism subjectively, which is inaccurate and time-consuming., by means of machine learning, the prediction of employees' absenteeism can be more objective and efficient. In this paper, we used the data provided by UCI machine learning database, which was created with records of absenteeism at work of a courier company in Brazil, to build absenteeism prediction model. We first conduct descriptive statistical analysis, and then employ four classical machine learning models to solve the problem. The integrated learning algorithm has the highest accuracy, which reaches 52% on the test set.