<p>Debris flow is a catastrophic geomorphic process that poses significant threats to human lives and infrastructures. An accurate prediction of entrainment growth rates is crucial for understanding debris flow dynamics and mitigating associated hazards. Traditional methods rely on empirical models and physically based simulations, but these approaches have limitations in capturing the complex and site-specific interactions governing entrainment processes. In this study, a deep neural network (DNN) model was developed to predict entrainment growth rates in various terrain conditions, utilizing a dataset from 54 debris flow sites across South Korea. The proposed model incorporates multiple geomorphological, hydrological, and geotechnical variables to improve the accuracy of entrainment growth. A comparative analysis was conducted against the regression model proposed by Park et al. (2024). The evaluation was performed using debris flow events from eight sites in South Korea, including five sites in Yecheon and three sites in Gwangju. Performance metrics including mean absolute error (MAE), root mean squared error (RMSE), and coefficient of determination (<i>R</i>²) were used to assess the accuracy of both models. The results demonstrate that the proposed DNN model outperformed the regression model, reducing MAE by 36%, RMSE by 25%, and improving <i>R</i>² from 0.68 to 0.79. The proposed DNN model showed higher accuracy than the regression model, effectively adapting to diverse terrains and environmental conditions within the tested South Korean cases. This study highlights the potential of DNN approaches in debris flow hazard assessment and suggests that the proposed model can serve as a promising predictive framework for estimating entrainment growth rate within the tested conditions.</p>