Driving behavior profiling has important relevance in many driving applications. For instance, car insurance companies have been recently applying a new insurance paradigm in which a driver's insurance premium is adapted based on realtime driving behavior. Driver profiling process is composed of two sub processes. The first is the detection of certain driving behaviors by acquiring data from onboard devices such as smartphones and OBDII units, whereas the second is the scoring process in which the detected behaviors are used to measure the actual driving risk. The scoring process has been viewed as an intricate problem due to the lack of reliable and large-scale datasets that can provide statistically trustworthy insights. This paper presents a data-driven approach for calculating a driver's risk score by utilizing the SHRP2 naturalistic driving dataset, which is the largest dataset of its kind to date. Two machine learning algorithms, which are support vector regression (SVR) and decision tree regression (DTR) are trained to reflect a driver's score. Driver's score is quantified in terms of the additive inverse of the predicted risk probability. After data filtering and preprocessing, models are trained using thirteen predictors, which represent twelve unique driving behaviors and the total driving time per driver. Validation results show that risk probability can be accurately predicted using the proposed models.