Motorbike Accident Severity Prediction Using Machine Learning Algorithms

Abstract

Motorbike accident is most dangerous among all accident on road. Many people dies because of bike accident. Motorbike accident damages both public and private property. A victim’s family have to spent a big amount of money for the treatment of the biker. Sometimes their family falls into financial crisis to arrange money for treatment of the victim. Bikers need to be aware about accident and should avoid the facts that causes accident. In this study different machine learning algorithm has been employed to predict the severity of the motorbike accident. Then we collect data depending on those criteria, such as speeding, overtaking, turning, bike fitness issues, speed-breakers without signs, unsafe lane changes, talking with a passenger, and highways without road dividers, among others. We only collect information from bikers who have been in an accident. We processed all of the data once it was collected and developed a processed dataset. On our processed dataset, we used machine learning techniques. Machine learning has been employed in various prediction and detection systems since their inception. Random Forest, Multilayer Perception (MLP), Decision Tree, Logistic Regression, k-Nearest Neighbors (KNN), AdaBoost, GNB, SVM with RBF Kernel, Linear SVC and Gradient Boosting are just a few of the techniques we utilize. MLP offered the greatest results in terms of accuracy. It showed an accuracy of 83.10%. And again MLP performs better in terms of sensitivity, specificity, F1-Score and precision.