# Retention Is All You Need

**Authors:** Karishma Mohiuddin, Mirza Ariful Alam, Mirza Mohtashim Alam, Pascal Welke, Michael Martin, Jens Lehmann, Sahar Vahdati  
**arXiv:** [2304.03103](https://arxiv.org/abs/2304.03103)  
**Date:** 2023-04-06  
**Summary page:** https://www.summarizepaper.com/en/paper/2304.03103v2/

## Key points

- HR-DSS approach proposed to address high attrition and turnover rates in organizations
- Utilizes explainable AI and machine learning models
- Experiments conducted with eight machine learning models
- Best-performing model processed using SHAP explainability process
- SHAP values generated natural language explanations for HR decision-making
- "What-if-analysis" conducted to observe plausible causes for attrition on an individual employee basis
- Adjusting dominant features of each individual can lead to informative business decisions and shift employee attrition towards retention
- Provides comprehensive solution for organizations to understand and address attrition challenges through interpretable predictions and actionable insights derived from machine learning models

## Summary

This paper proposes the HR-DSS approach to address high attrition and turnover rates in organizations. The system utilizes explainable AI and machine learning models to assist HR departments in interpreting predictions related to employee attrition. Experiments conducted employed eight machine learning models with the best-performing model further processed using the SHAP explainability process. The SHAP values generated natural language explanations for HR decision-making. Additionally, a "What-if-analysis" was conducted to observe plausible causes for attrition on an individual employee basis. Results demonstrate that by adjusting dominant features of each individual, informative business decisions can be made to shift employee attrition towards retention. This approach provides a comprehensive solution for organizations to understand and address their attrition challenges through interpretable predictions and actionable insights derived from machine learning models.