Driving behavior have shown to have an impact on proneness to vehicular accidents and fuel economy making it very interesting for insurance companies and fleet owners. By analyzing GPS data from drivers it is possible to evaluate and group drivers based on their driver behavior, measured in terms of acceleration, jerk, lateral acceleration, and wobble. To fairly group and evaluate driving behavior we introduce the term Hot Path, which is heavily traversed paths in a road network. Using Hot Paths, as a common denominator, we utilize k-means and DBSCAN along with the dimensionality reduction technique t-SNE to cluster driving behavior based on the observed data for each Hot Path. This paper presents a framework for generating Hot Paths for a large dataset of GPS trajectories using a novel variant of the Apriori algorithm. The framework is afterwards able to create meaningful clusters based on the observed data for the Hot Paths giving a user the tools to more easily evaluate and compare drivers using map matched GPS data. Through experimentation and evaluation using various cluster scoring methods we show that the framework is able to efficiently and effectively handle large datasets and find meaningful clusters in the data, e.g. representing calm and aggressive driving behavior.