This study explores the efficacy of using machine learning classifiers in traffic conflict identification. Quantitative conflict identification methods are largely designed through observation of motorized vehicles only, and can report erroneous results when applied to non-motorized modes.
Through video recordings of Bloor Street before and after an installation of bicycle lanes, the use of conflict analysis in identifying hazardous behaviour on multimodal streets is affirmed, and a dataset of conflict and non-conflict events is constructed. Six machine learning classifiers are trained on the dataset: three classifiers were trained using only conflict indicators, and three classifiers were trained using the full set of explanatory variables.
Machine learning techniques were found to be more effective in conflict identification than traditional threshold-based identification techniques, and user mode, speed, and acceleration may influence interaction severity. Combining machine learning with the results of automated video processing has particular promise in the field of conflict analysis.