In the context of autonomous driving and intelligent traffic systems, understanding the driving style of a driver is crucial for improving the interaction and adaptation of systems, thereby enhancing the adaptability and safety of autonomous driving systems. In this study, based on the HighD dataset, which provides data on vehicle driving states, we applied Principal Component Analysis (PCA) to reduce the dimensionality of 19 feature parameters that characterize driving style. These parameters include the average, maximum, minimum, and standard deviation of longitudinal and lateral velocities, as well as longitudinal and lateral accelerations, minimum time headway, minimum space headway, and minimum collision time. Subsequently, the extracted principal component data were clustered into three categories using the K-means clustering algorithm, classifying drivers into aggressive, normal, and cautious types. The clustering results indicated that aggressive drivers tend to accelerate or decelerate sharply, while general drivers pursue stability during driving, while cautious drivers tend to accelerate slowly during driving, which verifies that the clustering results are in line with reality.