@ARTICLE{26543117_316258389_2019, author = {Leandro Costa and Ricardo Ramos and Sérgio Moro}, keywords = {, absenteeism, human resources, public administrationdata mining}, title = {Anticipating Next Public Administration Employee’s Absence Duration}, journal = {Public Administration Issues}, year = {2019}, number = {6}, pages = {23-40}, url = {https://vgmu.hse.ru/en/2019--6/316258389.html}, publisher = {}, abstract = {Absenteeism affects state-owned companies who are obliged to undertake strategies to prevent it, be efficient and conduct effective human resource (HR) management. This paper aims to understand the reasons for Public Administration Employees’ (PAE) absenteeism and predict future employee absences. Data from 17,600 PAE from seven public databases regarding their 2016 absences was collected, and the Recency, Frequency and Monetary (RFM) and Support Vector Machine (SVM) algorithm was used for modeling the absence duration, backed up with a 10-fold cross-validation scheme. Results revealed that the worker profile is less relevant than the absence characteristics. The most concerning employee profile was uncovered, and a set of scenarios is provided regarding the expected days of absence in the future for each scenario. The veracity of the absence motives could not be proven and thus are totally reliable. In addition, the number of records of one absence day was disproportionate to the other records. The findings are of value to the Human Capital Management department in order to support their decisions regarding the allocation of workers and productivity management and use these valuable insights in the recruitment process. Until now, little has been known concerning the characteristics that affect PAE absenteeism, therefore enriching the necessity for further understanding of this matter in this particular.}, annote = {Absenteeism affects state-owned companies who are obliged to undertake strategies to prevent it, be efficient and conduct effective human resource (HR) management. This paper aims to understand the reasons for Public Administration Employees’ (PAE) absenteeism and predict future employee absences. Data from 17,600 PAE from seven public databases regarding their 2016 absences was collected, and the Recency, Frequency and Monetary (RFM) and Support Vector Machine (SVM) algorithm was used for modeling the absence duration, backed up with a 10-fold cross-validation scheme. Results revealed that the worker profile is less relevant than the absence characteristics. The most concerning employee profile was uncovered, and a set of scenarios is provided regarding the expected days of absence in the future for each scenario. The veracity of the absence motives could not be proven and thus are totally reliable. In addition, the number of records of one absence day was disproportionate to the other records. The findings are of value to the Human Capital Management department in order to support their decisions regarding the allocation of workers and productivity management and use these valuable insights in the recruitment process. Until now, little has been known concerning the characteristics that affect PAE absenteeism, therefore enriching the necessity for further understanding of this matter in this particular.} }