MODELING OF VEHICLE ACCIDENT PREVENTION AND CONTROL SYSTEM USING MACHINE LEARNING TECHNIQUE

Subject Area: CONTROL SYSTEM ENGINEERING

This paper presents the modeling of vehicle accident prevention and control system using machine learning technique. The study revealed from literatures that solution has not been obtained for this problem considering tricycle vehicles which has evolved recently as the major means of transport in Nigeria. This problem was addressed using Feed Forward Neural Network (FFNN), rule based and data collected from the ministry of transportation. The FFNN was used to develop the tricycle detection model, while the rule based was used to develop the accident detection and control system using the tricycle detection model and safe distance standard between vehicles according to the Federal Road Safety Corp (FRSC) which is not less than 1.3meters. The model was implemented with Simulink platform and then evaluated. The result of the accident detection model was evaluated with Means Square Error (MSE), Receiver Operator Characteristics (ROC) curve and confusion matrix. The MSE 3.0512e-10, accuracy is 98.1% and ROC is 0.9807. The model was compared with other sophisticated accident detection and control model and the result showed that the new system has a percentage improvement of 5.1%. Keywords: Vehicle; Accident; Neural Network; Tricycle; Receiver Operator Characteristic