Traffic status detection on arterial roads is challenging because of the complexity of urban traffic and the limited coverage and high deployment cost of traffic detectors. Ubiquitous mobile phones and data generated from the events of these mobile devices provide a promising approach for traffic status detection. This paper proposes a novel approach solely using cellular event data without the cellphone GPS information to detect traffic status on arterial roads. Different from the conventional methods, the proposed approach uses features derived only from cellular data to estimate traffic status, not requiring any cellphone location information. Both handoff (HO) and location update (LU) events generated at each cellular station were extracted from the original data to form a candidate feature set. A feature selection method based on joint mutual information (JMI) was used to select features to cover the maximum information, which can resolve issues such as loss of useful information caused by conventional feature selection techniques. A support vector machine (SVM) algorithm was then employed to model the relationship between the selected features and traffic status (low, medium, and high-traffic). Finally, the proposed method was validated by both a field experiment in Taicang, China with 1-hour-time-interval samples and a simulation experiment on VISSIM with 5-minute-time-interval samples. This study provides a new perspective for traffic status detection which may help design strategies for traffic management and route navigation to improve traveling efficiency, especially for the cities lack of traffic surveillance devices.