Machine Learning Algorithms as a Tool for Improving Road Safety

The chapter relates to the development of methods’ comparative analysis. It investigates predictive machine learning techniques for the aim of the road safety improving. This study provides baseline analysis as a method for the evaluation of the model better quality. A baseline method allows us to compare different predictive models between each other and solve the problem of finding better way for an analysis. The analysis in this study is carried out on the basis of the machine learning techniques and the crash data in Saint-Petersburg. The results show a significant impact of oversampling in the slight class of severity level. Such peculiarity of the dataset influences on the reliability of models’ results. The current analysis confirms that baseline is effective method of different methods’ comparison, but also the crash data has the pattern of unbalancing that can cause of obtaining the proper results. Therefore, further direction of the research is applying different oversampling methods with machine learning predictive models.