| Abstract : | Employee attrition may be genuine concern in information-based organizations. When representatives take off an organization, they carry essential inferred information that is regularly the source of competitive advantage for the firms. For an organization to persistently have the next competitive advantage over this competition, it ought to make it an obligation to minimize employee steady loss. Numerous components might be responsible for that (for case: social, budgetary, ramp downs, managerial issues etc.). Each organization has a certain way to behave its employees and assure their satisfaction with all aspects. However, frequently no measures are taken as regards the fulfillment rate. Therefore, in numerous cases, personnel resigned suddenly without an apparent justification. This proposal investigates the effectiveness of survival analysis & machine learning algorithms in forecasting employee attrition. In specific, the current thesis mentions the predictive performance of the Cox Proportional Hazards - Model and some machine learning algorithms such as DeepSurv, Random Survival Forest, max -out. To illustrate the impact of the above to employee turnover, IBM’s synthetic workforce data from Kaggle were used where the Random Forest algorithm was applied to predict the future turnover of employees.
H παρούσα Διπλωματική διερευνά την αποτελεσματικότητα της ανάλυσης επιβίωσης και των αλγόριθμων μηχανικής μάθησης στην πρόβλεψη της αποχώρησης των εργαζομένων. Πιο συγκεκριμένα στην εν λόγω εργασία αναφέρεται η προβλεπτική απόδοση του στατιστικού μοντέλου του Cox Proportional Hazards - Model και ορισμένων αλγορίθμων μηχανικής μάθησης όπως Random Survival Forest, max -out. Για να καταδειχθεί η επίδραση των παραπάνω στην κινητικότητα των εργαζομένων, χρησιμοποιήθηκαν συνθετικά δεδομένα εργατικού δυναμικού της IBM από το Kaggle, όπου εφαρμόστηκε ο αλγόριθμος Random Forest για την πρόβλεψη της μελλοντικής κινητικότητας των εργαζομένων
|
|---|
Copyright © 2013 Library AUEB. All rights reserved.