With the wider availability of sensor technology through easily affordable sensor devices, many structural health monitoring (SHM) programs and wind field monitoring networks have been established to better understand the different meteorological phenomena and their effects on structures. A deluge of high-dimensional, heterogeneous data is continuously generated by these monitoring networks. Efficient data processing from massive volumes of possibly uncertain data has become a daunting task. This dissertation demonstrates the value added by combining machine learning techniques and sensor data that will help in solving many of the complex problems which otherwise cannot be resolved by conventional analysis techniques. Various machine learning frameworks have been developed in this study to better understand extreme wind characteristics and the performance of civil infrastructure under winds.Three types of long-term monitoring networks are considered in this study and three machine learning frameworks are developed and applied to each of these networks. SmartSync, a structural health monitoring system installed on Burj Khalifa, the world's tallest building, is used to obtain data about the structure's dynamic behavior. A generalized framework for cluster analysis is then developed to effectively identify and extract vortex-induced vibration of the pinnacle of Burj Khalifa and to estimate the crosswind fatigue damage of the pinnacle. Wind and Ports (WP) and Wind, Ports, and Sea (WPS) projects, that consist of a network of anemometers situated in various locations in ports of Italy and France are considered in this study. A relatively new time series representation named "Shapelet Transform" is introduced in combination with machine learning algorithms to autonomously identify thunderstorms from the data obtained through this wind monitoring network. Finally, an SHM system on the Sutong Yangtze River Bridge, a long-span cable-stayed bridge in China is also considered in this study. The shapelet-based machine learning framework is used to classify six different anomalous patterns in the wind-induced vibration data obtained from the SHM system. The contributions of this work have established a framework for applying machine learning techniques to big data from long-term monitoring networks. Each of these will facilitate automation while mining massive databases and yield a practical event detection procedure that requires minimal domain expertise.