Antarctica’s coastal “fast ice” is vital for weather, wildlife, and scientific research at places like Jang Bogo Station of South Korea. However, it's shrinking rapidly. Currently, predicting future fast ice levels is difficult. Our study developed a new machine learning model to forecast how much fast ice will be present. We used data on air temperature, wind patterns, and satellite data derived open water areas. We discovered that considering weather conditions from previous months significantly improved prediction accuracy, meaning past temperatures and winds are key indicators. Specifically, air temperature one month prior and wind patterns two months prior were most influential. This model represents a major advancement in understanding fast ice dynamics. It offers valuable insights for adapting to the changing Antarctic environment and supporting logistical planning as the climate evolves, helping scientists work more effectively in this crucial region.