EFFECTIVE IMPLEMENTATION OF VEHICULAR ACCIDENT DETECTION AND PREVENTION USING MACHINE LEARNING TECHNIQUE

Abstract

This paper analyzes the use of rule-based integrated machine learning technique to model inter-vehicular accident prevention and control system. It has been noted that tricycle road movements are the major cause of road accident, even though they have dominated the means of transportation and logistics in many developing countries, especially in Nigeria. To solve this problem of road accident, data of tricycles were collected from the Anambra State Ministry of Transport, Awka; and then trained with machine learning algorithm to generate the accident detection model. The rule-based optimization was developed from the information collected from the Federal Road Safety Corp (FRSC) on the standard of inter vehicle distance and then used to develop the control model. The model was implemented with Simulink and evaluated. The result when tested and validated showed that the accident detection accuracy is 98.1%; Mean Square Error (MSE) is 3.0512e-10 and ROC is 0.9807. When compared with other models trained with similar data type, the result showed that the Feed Forward Neural Network (FFNN) developed was better and more accurate with a percentage improvement of 5.1%.