Abstract
This study conducted an exploratory analysis of bicycle crash data from Great Britain with the aim of identifying the key variables that influence the classification of such incidents. It also analysed data on a range of factors that may contribute to bicycle crashes, including the age of the cyclist, lighting conditions, weather conditions, road types, road conditions, and speed limits. Results indicated that these variables are among the most significant predictors of bicycle crashes, with road conditions, time of day, and lighting conditions being particularly vital factors. In addition, the study sought to compare the efficacy of different machine learning and deep learning models in predicting the severity of such incidents. Results indicated that these models demonstrated poor performance in predicting the severity of bicycle crashes. As a result, a hybrid model that combines the K-Nearest Neighbor and eXtreme Gradient Boosting algorithms was developed to improve accuracy. The hybrid model outperformed all other models, achieving an accuracy rate of 83.56%. The study, additionally, has put forward several recommendations, including the mandatory use of reflective clothing and the installation of Intelligent Transportation Systems (ITS) to enhance the safety of cyclists.