An Accurate Model for Hurricane Trajectory Prediction

Hurricanes have caused billions of dollars' worth of damage due to property destruction, personal danger, and civil unrest. This catastrophic impact of hurricanes has motivated extensive research in hurricane forecasting, within which is the subfield of hurricane trajectory prediction. Traditionally, meteorological analysis was used in determining the future path of a given hurricane. In this paper, we present an alternative approach. Specifically, we measure realistic hurricane trajectories by their haversine distances. We train our prediction model exclusively with data from years 1950 to 2000. Moreover, we weigh the training data to favor recent hurricane trends when determining a trajectory prediction. Evaluation results show that our model led to a prediction-correctness ratio up to 10.0% higher than the current state-of-the-art, which is at 75.0%.