Analysis of Typhoon Path for the Korean Peninsula using Machine Learning Algorithms

This research enhances typhoon path prediction by integrating advanced machine learning with extensive meteorological data, focusing on the Korean Peninsula. Utilizing a novel two-step modeling approach, it predicts if typhoons will cross latitude 33 degrees and their paths thereafter, categorizing them into four types based on their trajectories relative to the Peninsula. Models were assessed using data from 1979 to 2023, employing variable importance and partial dependence plots to decipher the decision-making processes. Findings show that longitude and meteorological factors like air and skin temperature significantly affect typhoon paths. The study sets a new standard in typhoon forecasting by substituting high-dimensional spatio-temporal data with localized datasets and offers insights for disaster management, setting a new standard in typhoon forecasting.