Publicação
Human Resources Analytics for Turnover Prediction: Insights from an Outsourcing Software Company
| Resumo: | Retaining top talent in companies is critical. This study explores the application of Machine Learning (ML) techniques to predict employee turnover and identify key turnover motivators within an IT outsourcing company. Using the CRISP-DM methodology, the study follows a structured approach to business understanding, data preparation, model training, and evaluation. A total of ten ML models were developed, with an Artificial Neural Network (ANN) achieving the highest predictive performance, as measured by F1 scores. Model explainability was addressed through SHAP (SHapley Additive exPlanations) values, enabling the identification of the most influential factors in predicting turnover. The results reveal that project duration, absence days, time to change contract terms, and commissions were among the most significant predictors of employee turnover. All models demonstrated strong performance, highlighting the feasibility of using ML in Human Resource Analytics (HRA) for this purpose. This study provides insights for companies seeking to proactively manage turnover by highlighting the key drivers influencing employees’ decisions to leave. |
|---|---|
| Autores principais: | Carvalho, Mariana Antunes Figueiredo |
| Assunto: | Human Resource Analytics People Analytics Employee Turnover CRISP-DM Predictive Analytics SDG 8 - Decent work and economic growth SDG 9 - Industry, innovation and infrastructure |
| Ano: | 2025 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso embargado |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | Retaining top talent in companies is critical. This study explores the application of Machine Learning (ML) techniques to predict employee turnover and identify key turnover motivators within an IT outsourcing company. Using the CRISP-DM methodology, the study follows a structured approach to business understanding, data preparation, model training, and evaluation. A total of ten ML models were developed, with an Artificial Neural Network (ANN) achieving the highest predictive performance, as measured by F1 scores. Model explainability was addressed through SHAP (SHapley Additive exPlanations) values, enabling the identification of the most influential factors in predicting turnover. The results reveal that project duration, absence days, time to change contract terms, and commissions were among the most significant predictors of employee turnover. All models demonstrated strong performance, highlighting the feasibility of using ML in Human Resource Analytics (HRA) for this purpose. This study provides insights for companies seeking to proactively manage turnover by highlighting the key drivers influencing employees’ decisions to leave. |
|---|