Predicting Oversea Success of K-Content on OTT Platforms: Based on Machine Learning and SHAP Analysis

K-content, including K-dramas, has been receiving significant global attention in recent years. This study aims to predict the overseas success of K-dramas on OTT platforms and to identify the key factors contributing to their popularity. To achieve this, data on K-dramas aired abroad were collected from five sources: Kaggle, FlixPatrol, Naver, Prettyscale, and Instagram. Among them, 279 dramas with available ranking information were selected to build prediction models using three machine learning algorithms: Random Forest, XGBoost, and CatBoost. In addition to constructing a model targeting the global market, we developed two more models by segmenting the data into English-speaking and non-English-speaking countries, considering cultural differences. Through feature importance and SHAP analysis, we found notable differences in key factors—such as genre preferences, actors’ appearance, and release timing—between the English-speaking and non-English-speaking markets. The findings of this study offer valuable insights for K-content producers and OTT platforms in formulating market-specific strategies and enhancing their chances of success in the global content market.