Predicción de la rotación de personal y optimización de la estructura de retención en un outsourcing delivery center usando Machine Learning y Programación Multiobjetivo
No hay miniatura disponible
Fecha
2024-07-19
Título de la revista
ISSN de la revista
Título del volumen
Editor
Pontificia Universidad Católica del Perú
DOI
Resumen
La rotación de los empleados se ha convertido en un foco de investigación del área de
Recursos Humanos porque tiene efectos significativos en el performance de las
organizaciones independientemente de la geografía, tamaño de la empresa o sector. La
rotación de personal afecta a la empresa como negocio y como cultura laboral, muchos creen
que sus efectos son relativamente fáciles de medir: el costo incurrido en contratar y capacitar
nuevo personal, pero una rotación de personal elevada supone costos ocultos como la pérdida
de confianza en el empleador, ambiente laboral dañino, además que permite fugas de
información y poco sentido de permanencia.
En este sentido, el uso de Machine Learning para predecir la probabilidad de que un empleado
renuncie a su trabajo podría aumentar en gran medida la capacidad del departamento de
Recursos Humanos para intervenir a tiempo y proporcionar un enfoque mitigador a esta
situación.
Este estudio se realiza con el objetivo de comparar el rendimiento de las técnicas de
aprendizaje automático, entre ellos XGBoosting, decisión tree, random forest, KNN, SVM,
logistic regression, y LGBM y seleccionar el mejor modelo que busque predecir la
permanencia de candidatos en su primer año de labores en posiciones técnicas, de
planeamiento y estratégicas en una empresa outsourcing peruana. Todo esto se realizará bajo
el modelo estándar abierto CRISP-DM, un método probado para orientar trabajos de minería
de datos. Finalmente, tomando como input los resultados del clasificador seleccionado, se
construirá un modelo de optimización bi-objetivo no lineal que buscará minimizar los costos
tras aplicar estrategias de retención y reducir la brecha salarial entre el salario actual y el del
mercado para minimizar la tasa de rotación. El estudio ayudará a que la gerencia pueda
adoptar técnicas de retención de personal basada en los atributos que impactan más en la
decisión de un empleado de renunciar voluntariamente a una empresa.
Employee turnover has become a research focus in the Human Resources area because has significant effects on the performance of organizations regardless of geography, company size or sector. Staff turnover affects the company as a business and as a work culture, many managers believe that its effects are easy to measure: the cost incurred in hiring and training new staff, but high staff turnover implies hidden costs such as loss of confidence in the employee, a harmful work environment, in addition to allowing information leaks and little sense of permanence. In this sense, the use of Machine Learning to predict the probability that an employee will quit their job could greatly increase the ability of the Human Resources department to intervene in time and provide a mitigating approach to this situation. This study is carried out to compare the performance of machine learning techniques, including XGBoosting, decision tree, random forest, KNN, SVM, logistic regression, and LGBM, to select the best model that seeks to predict the retention of candidates in their first year of work in technical, planning, and strategic positions in a Peruvian outsourcing company. All of this will be done under the open standard CRISP-DM model, a proven method for guiding data mining work. Finally, taking as input the results of the selected classifier, a bi-objective nonlinear optimization model will be built to minimize costs after applying retention strategies and reduce the wage gap between the current salary and the market salary, aiming to minimize the attrition rate. The study will help management to adopt retention techniques based on the attributes that most impact an employee's decision to voluntarily resign from a company.
Employee turnover has become a research focus in the Human Resources area because has significant effects on the performance of organizations regardless of geography, company size or sector. Staff turnover affects the company as a business and as a work culture, many managers believe that its effects are easy to measure: the cost incurred in hiring and training new staff, but high staff turnover implies hidden costs such as loss of confidence in the employee, a harmful work environment, in addition to allowing information leaks and little sense of permanence. In this sense, the use of Machine Learning to predict the probability that an employee will quit their job could greatly increase the ability of the Human Resources department to intervene in time and provide a mitigating approach to this situation. This study is carried out to compare the performance of machine learning techniques, including XGBoosting, decision tree, random forest, KNN, SVM, logistic regression, and LGBM, to select the best model that seeks to predict the retention of candidates in their first year of work in technical, planning, and strategic positions in a Peruvian outsourcing company. All of this will be done under the open standard CRISP-DM model, a proven method for guiding data mining work. Finally, taking as input the results of the selected classifier, a bi-objective nonlinear optimization model will be built to minimize costs after applying retention strategies and reduce the wage gap between the current salary and the market salary, aiming to minimize the attrition rate. The study will help management to adopt retention techniques based on the attributes that most impact an employee's decision to voluntarily resign from a company.
Descripción
Palabras clave
Rotación de personal, Cultura organizacional, Aprendizaje automático