Publicação
Predicting Employee Turnover in SME's: A Data-Driven Approach for Workplace Retention
| Resumo: | This thesis addresses the difficulty of predicting employee turnover in small and medium-sized organizations (SMEs) using a data-driven, machine learning-based strategy. The study is driven by the particular setting of small and medium-sized enterprises (SMEs), which frequently lack the resources, data volume, and established HR procedures that exist in large businesses. The study uses anonymized longitudinal data from a Portuguese SME in the food manufacturing sector to identify the main factors influencing turnover. A Comprehensive literature review guides the selection of machine learning models and addresses the challenges posed by small, imbalanced datasets, common in SME settings. The results show that ensemble tree models regularly outperform alternative approaches in terms of accuracy and interpretability. Notably, the study’s key conclusion is that turnover is best explained by a mix of demographic (age, entrance age), financial (monthly income, bonuses), and organizational (section, distance from company) characteristics, rather than a single predictor. In this investigation, synthetic data augmentation with CTGAN did not consistently improve model performance. While it occasionally enhanced data diversity, it had some negative effects on both ensemble and simpler classifiers in the SME scenario. This thesis suggests that, with rigorous preprocessing and careful model selection, predictive analytics can provide actionable insights for SME leaders, allowing for focused retention measures that could cut turnover by at least 10%. The study was focused on a single-company scope, a small sample size, and the exclusion of qualitative characteristics, which is clearly a limitation. Future research prospects incorporating bigger, multi-company datasets and the incorporation of qualitative employee feedback, being off-boarding interviews a must. Overall, this study reveals machine learning's practical potential to improve HR decision-making and workplace stability in data-constrained SME situations. |
|---|---|
| Autores principais: | Jardim, Sara Barreto |
| Assunto: | Human Resources Employee Turnover Employee Retention Predictive Modeling Small and Medium-Enterprises SDG 8 - Decent work and economic growth SDG 9 - Industry, innovation and infrastructure SDG 10 - Reduced inequalities |
| Ano: | 2025 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | This thesis addresses the difficulty of predicting employee turnover in small and medium-sized organizations (SMEs) using a data-driven, machine learning-based strategy. The study is driven by the particular setting of small and medium-sized enterprises (SMEs), which frequently lack the resources, data volume, and established HR procedures that exist in large businesses. The study uses anonymized longitudinal data from a Portuguese SME in the food manufacturing sector to identify the main factors influencing turnover. A Comprehensive literature review guides the selection of machine learning models and addresses the challenges posed by small, imbalanced datasets, common in SME settings. The results show that ensemble tree models regularly outperform alternative approaches in terms of accuracy and interpretability. Notably, the study’s key conclusion is that turnover is best explained by a mix of demographic (age, entrance age), financial (monthly income, bonuses), and organizational (section, distance from company) characteristics, rather than a single predictor. In this investigation, synthetic data augmentation with CTGAN did not consistently improve model performance. While it occasionally enhanced data diversity, it had some negative effects on both ensemble and simpler classifiers in the SME scenario. This thesis suggests that, with rigorous preprocessing and careful model selection, predictive analytics can provide actionable insights for SME leaders, allowing for focused retention measures that could cut turnover by at least 10%. The study was focused on a single-company scope, a small sample size, and the exclusion of qualitative characteristics, which is clearly a limitation. Future research prospects incorporating bigger, multi-company datasets and the incorporation of qualitative employee feedback, being off-boarding interviews a must. Overall, this study reveals machine learning's practical potential to improve HR decision-making and workplace stability in data-constrained SME situations. |
|---|