Internet of things connects any object to another and communicate with them using the most performed routing protocol. But in vehicular networks, topology and communication links frequently change due to the high mobility of vehicles. So, the key challenge of our work is to choose the best routing protocol using machine learning algorithms. When choosing routing protocol, most research focuses on the improvement of the performance of specific routing protocol using one machine learning algorithm. In this paper, we propose a solution in order to find the best routing protocol in such critical condition using machine learning algorithms in order to maximize the precision of the true positive rate. After the use of a specific algorithms such as Artificial Neural Network, Random Forest and Naive Bayes, we found that the last one is the best algorithm when it have all the true positive rate with a precision equal to 0.9987 to select the best routing protocol.