Vehicular Ad hoc Network (VANET) is superior to traditional Intelligent Transportation Systems (ITS) for its vehicle-to-x communication. VANET is established on the underlying information exchanges and broadcast by vehicles and infrastructures in the network. The driving status of vehicles on road can be got from the periodically broadcasts messages and the particular data can be stored to be exploited for special service. These obtained data are richer and more representative than those data collected by fixed sensors such as ground induction coil, floating car data or video analysis, which can be utilized to get the real-time road traffic knowledge in a way similar to big data technology. And the existing relevant approaches about traffic detection and prediction face the following disadvantages: (1) insufficient data utilized; (2) hand engineered in features; and (3) suboptimal learning method. As the big data about the real time traffic can be obtained through VANETs conveniently, this work thus proposed a road functionality detection model based on data aggregation in VANETs applying machine learning algorithm. The performance of the scheme is evaluated through extensive simulations. The simulation results show that the approach of fusing the information of vehicles on the road can accurately detect the dysfunctional road state and further locate on the specific malfunctioning position particularly for high dimensions.