A carregar...
Publicação
Employee Turnover Predictive Model: A model to predict who is more likely to leave a company
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.73 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
This master's thesis explores the predictive modeling of employee turnover within a company
in the Retail & Food Industry, using a data-driven approach to identify potential leavers and
understand the dynamics affecting their decisions. Employing machine learning techniques
such as Logistic Regression, Random Forest, and Neural Networks, the study focuses on
optimizing the prediction of employee turnover through sophisticated model selection and
hyperparameter tuning. The research began with an extensive data preparation phase, which
involved cleaning, normalization, and transformation of the dataset to ensure robustness and
relevancy for model training. This process also included the application of SMOTE (Synthetic
Minority Over-sampling Technique) to address class imbalance within the dataset, ensuring
that the predictive performance was not biased towards the majority class. Key features
influencing turnover, such as job satisfaction, management styles, compensation packages,
and career progression opportunities, were identified and engineered to enhance the
predictive performance of the models. Several models were evaluated, with the ensemble
approach integrating Random Forest, Gradient Boosting, and Neural Networks showing the
most promising results. This ensemble model, optimized for high precision in predicting
'Active' status without overfitting, achieved remarkable accuracy (95.1% to 95.8%), precision
for label 0 (non-leavers) up to 92.5%, and an ROC-AUC score demonstrating excellent
classification capabilities (up to 0.983). The refined models significantly outperformed initial
predictions, highlighting the effectiveness of the feature selection and machine learning
techniques employed. The findings suggest that the integrated approach can effectively
predict employee turnover, providing HR departments with a valuable tool for strategic
human resource planning. This predictive capability enables proactive interventions tailored
to mitigate turnover and enhance employee retention strategies. In conclusion, this thesis not
only demonstrates the applicability of advanced analytical techniques to real-world HR
challenges but also lays the groundwork for future research. It suggests exploring further
cross-industry applications, integration of additional data sources, and employing alternative
modeling techniques to expand the model's robustness and adaptability across different
organizational contexts.
Descrição
Dissertation presented as the partial requirement for obtaining a Master's degree in Information Management, specialization in Knowledge Management and Business Intelligence
Palavras-chave
Employee Termination Employee Turnover Predictive Modelling Machine Learning Human Resources Analytics Employee Retention SDG 8 - Decent work and economic growth