A Survey on Road Accident Prediction Techniques Based on Various Methodologies

Abstract

Since traffic accidents are a leading source of injury and death globally, there has been a lot of focus on developing more accurate methods of analysis and prediction in order to pinpoint the causes of these tragedies. Predicting traffic accidents is an effort to meet the problem of creating a safer transportation environment in order to save lives. The purpose of this study is to survey the current landscape of research into the use of convolutional neural networks, long short-term memory networks, and other deep learning architectures for the prediction of traffic accidents. In addition, the most popular data sources for predicting traffic accidents are compiled here and analyzed. Additionally, a categorization is recommended based on factors including its source and features, such as open data, measuring methods, onboard equipment, and social media data. In this section, we list and evaluate the many algorithms used to forecast traffic accidents, taking into account the data types for which each is most suitable, the accuracy of the findings, and the clarity with which they can be interpreted and studied. In order to further analyze the findings, the authors found that the best results were achieved by combining two or more analytic approaches. Many authors agree that using geospatial data, information from traffic volume, traffic statistics, video, sound, text, and sentiment from social media may improve the precision and accuracy of the analysis and predictions; this is one of the next challenges in road traffic forecasting.