See our Medium article for further details: click here
This repo contains two proof-of-concept machine learning classifiers, created by Saad Akbar and Keagan Rankin at University of Toronto, that classifies roads as high-stress or low-stress for cyclists based on a set of features. The classifier attempts to reduce the volume of work required in the popular LTS classification method (Furth et al. in 2016, Imani et al. 2019, Lin et al. 2021) the model was trained using open Toronto data and labels from the cited literature.
The final models are saved in all_roads_model.py and all_modes_model.py in the projsubmission folder. See the notebook final_models.ipynb in the same folder for an example of how to use the models. These models vary based on input features (see the files allmodes_train.csv and centrelinebike_train_spatial.csv in the train data folder for feature format). Models can be saved as a binary and used for prediction or improved training on data/labels from other cities.
NOTE: AS OF CURRENT VERSION, FEATURE REGIONS AND FEATURE CREATION FUNCTION add_regions() SHOULD BE DROPPED FROM THE MODEL IF APPLYING IT TO CITIES OTHER THAN TORONTO.
see dependencies.txt for dependencies
Contact authors for comments or suggestions.
Furth, P. G., Mekuria, M. C., & Nixon, H. (2016). Network connectivity for low-stress bicycling. Transportation Research Record, 2587(1), 41-49. DOI: https://doi.org/10.3141/2587-06.
Imani, A. F., Miller, E. J., & Saxe, S. (2019). Cycle accessibility and level of traffic stress: A case study of Toronto. Journal of Transport Geography, 80, 102496. DOI: https://doi.org/10.1016/j.jtrangeo.2019.102496.
Lin, B., Chan, T. C., & Saxe, S. (2021). The Impact of COVID-19 Cycling Infrastructure on Low-Stress Cycling Accessibility: A Case Study in the City of Toronto. Findings, 19069. DOI: https://doi.org/10.32866/001c.19069.