With the increase in road accidents in several countries, there is a need for an inexpensive and reliable system that can be retrofitted to any type of vehicle and can predict and report road accident events. This work aims to demonstrate an IoT system that uses in-built sensors of a smartphone to report road accidents accurately. NHTSA (National Highway Traffic Safety Administration) datasets have been used to train and test our accident prediction models. The prediction performance was found to be highly accurate with a MAPE of 2.34 percent and 3.49 percent respectively, for the datasets used. © 2019 IEEE.