This study developed a model to predict employee turnover intention using data from the 2022 Korean Labor & Income Panel Study (KLIPS) with 2471 participants. CopulaGAN and Isolation Forests were employed for data augmentation and variable importance. A logistic regression model using the augmented data achieved an accuracy of 0.80, precision of 0.60, recall of 0.72, and an F1-score of 0.65. Key variables included Job Satisfaction, Wage Satisfaction, Work Hours, Job Stability, and Job-Related Training. The study highlights the potential of these techniques for enhancing turnover prediction and aiding proactive HR strategies.

Abstract
1. Introduction
2. Methods
    2.1. Data collection
    2.2. Data Preprocessing
    2.3. Addressing Data Imbalance
    2.4. Detailed Steps of the Algorithm
    2.5. Developing and Evaluating the PredictionModel
3. Results
4. Discussion
5. Conclusion
References

• 변해원(Dept. of AI-Software, Inje University, South Korea) | Haewon Byeon Corresponding author