A study on improving turnover intention forecasting by solving imbalanced data problems: focusing on SMOTE and generative adversarial networks

Abstract This study aims to improve the accuracy of forecasting turnover intention new college graduates by solving imbalance data problem. For this purpose, from Korea Employment Information Service's Job Mobility Survey (Graduates Occupations Survey: GOMS) for were used. includes various items such as intention, personal characteristics, and job characteristics graduates, class ratio is imbalanced. problem, synthetic minority over-sampling technique (SMOTE) generative adversarial networks (GAN) used balance variables examine improvement prediction accuracy. After deriving factors affecting referring previous studies, a model was constructed, model's analyzed reflecting each data. As result analysis, highest predictive found in balanced through rather than imbalanced original SMOTE. The academic implication that first, diversity sampling methods presented expanding applying GAN, which are widely unstructured fields images images, structured business administration study. Second, two refining processes performed on generated using suggest method only corresponding more class. practical it suggested plan predict early establishment public machine learning.