A transformer-based deep learning framework to predict employee attrition.

Area of Science:

• Business Analytics
• Artificial Intelligence
• Machine Learning

Background:

• Employee attrition negatively impacts business profit management.
• Predictive modeling offers a solution to mitigate costs associated with employee turnover.
• Existing models lack real-world evaluation and integration into decision support systems.

Purpose of the Study:

• To evaluate Transformer-based neural networks for employee attrition prediction.
• To assess the integration of predictive models into business decision support systems.
• To determine the impact of accurate attrition prediction on strategic business decisions.

Main Methods:

• Implementation of a Transformer-based neural network model.
• Utilizing contextual embeddings for adaptation to tabular data.
• Application to the IBM HR Employee Attrition dataset.

Main Results:

• The Transformer model demonstrated significantly improved prediction efficiency over state-of-the-art models.
• The study validated the effectiveness of the proposed computational technique.
• Deep learning, especially Transformer networks, shows promise for tabular and unbalanced data.

Conclusions:

• Transformer-based neural networks are highly effective for predicting employee attrition.
• These models can be integrated into decision support systems for strategic business planning.
• Deep learning offers a powerful approach for analyzing complex, real-world business data.