Extreme weather events, particularly high wind speeds, present a serious risk to critical infrastructures worldwide, often resulting in severe damage and substantial economic losses. Transportation infrastructure, especially bridges, is particularly vulnerable, where high winds can cause vehicles to overturn, creating life-threatening hazards. This research investigates the dynamics of extreme wind events to develop strategies aimed at mitigating their impacts and enhancing infrastructure resilience. The study examines the interplay between wind directionality and the imbalance ratio of extreme wind speeds to improve predictive capabilities, which are crucial for evaluating risks such as vehicle overturning on bridges. A long short-term memory (LSTM) model is employed to predict extreme wind speeds, complemented by traditional extreme value analysis to identify these events across various directional sectors. The model is evaluated using two datasets from sensors in Los Angeles and Chicago. To accomplish this, traditional extreme value analysis is used to identify and extract extreme wind events across the different directional sectors from the two locations, setting the threshold for the classification. The findings of this research offer valuable insights into the characteristics of extreme winds, aiming to advance the development of more effective prediction models and safety measures to protect essential infrastructure from their potentially catastrophic effects.