Predicting Number of Accidents and Black Spot of a Route Using Genetic Algorithm: Improving Traffic Accident Prediction with Genetic Algorithm-based Method

Abstract

Because of its importance in saving human lives, traffic accident prediction is a motor vehicletraffic challenge. There are various studies in the literature that use artificial neural networks (ANNs),support vector machines (SVMs), decision trees (DTs), and other categorization approaches to predictthe severity of traffic accidents. Indeed, the fundamental shortcoming of ANNs and SVMs is their lack ofhuman interpretation, whereas the main disadvantage of traditional DTs like C4.5, ID3, and CART is theirlow accuracy. To solve these flaws, we present a Genetic Algorithm-based method to predict trafficaccidents based on user preferences instead of traditional DTs in this review.We customised a geneticalgorithm, to optimise and find rules based on Support, Confidence, and Comprehensibility metrics inthe suggested method. The suggested method's goal is to provide facilities for users, such as trafficcops, road and transportation engineers, to make use of their knowledge while balancing all of thecompeting objectives. A traffic accident data set of accidents in rural and urban roadways in TehranProvince, Iran, was used to assess the suggested technique during a five-year period (2008–2013).According to the evaluation results, the proposed technique outperforms classification methods such asANN, SVM, and traditional DTs in terms of classification metrics such as accuracy and rule performancemetrics such as support and confidence.