waze-romania

Using Waze data to better understand mobility issues in Romania.

In this repository, we aggregate traffic data from Waze in Cluj from individual points to hexagons (from 36 km^2 to 0.015km^2 in size) for vizualization and analysis of traffic alerts in the city for the years 2019 until 2022. We then run various machine learning algorithms to predict daily traffic alerts in distinct regions of Cluj given spatial and temporal variables.

Predictive Features

We incorporate spatial variables on hospital and school locations in the city and bus stop locations. Temporal variables included in our models are Covid-19 stringency index, precipitation, day of the week, month, and indicators for a home soccer occuring and whether it is a holiday.

Workflow

1. Read in and clean data from waze: clean_aggregate.ipynb
2. Read in and clean spatial data for the city of Cluj: Shape files.ipynb
3. Web-scraping and final data manipulation for modeling: modelling_data.ipynb
4. Running the final models: ML.ipynb