Challenge provided by Urbananalytica

Identifying patterns, explanatory factors, and prediction of irregular parking

All solutions point to parking infractions being more common on arterial roads and during big events.

As the population that lives, works, and visits cities increases, an increasing number of cars circulating on the streets of these cities, especially when public transportation offer is subpar. Consequently, parking capacity is under significant pressure, with offers seemingly failing to follow demand, which leads to the need for new solutions.

As a starting point, predicting irregular parking can help city administrative services to optimize parking inspection and dissuade possible irregular behavior from drivers. A consequence of this will also be less congestion since there will be minor blockage caused by cars in the streets, leading to lower CO₂ emissions.


Develop an explainable predictive model for irregular parking at street level and day.

United Nations SDG 

GOAL 11: Sustainable Cities and Communities

  • Target 11.7: Provide access to safe and inclusive green and public spaces


The following datasets were provided to the participants:

  • Parking tickets from 2017-2019 in Vancouver, by street and type of infraction, Open Data by the City of Vancouver.
  • Characteristics of public streets in Vancouver, Open Data by the City of Vancouver.


In addition to the provided datasets, more data was used by two of the teams. One team, which focused extensively on parking infraction causality derived from holidays and other events, gathered a vast amount of data related to bank holidays in Vancouver and major events such as concerts, strikes, and severe weather phenomena during the time comprised by the dataset.

Another team gathered open data about the weather in Vancouver to analyze if it could be a factor or predictor in the number of parking infractions.

Methods and Techniques

During exploratory data analysis (EDA), teams focused on understanding the characteristics of traffic violations. This included analyzing the location of the infraction, the number of infractions per day, the type of infraction, and even the weather conditions. Analyzing the location of the infraction involved encoding categorical variables. One team analyzed the data to determine its stationarity, seasonality, and trend using rolling statistics methods, such as the Dickey-Fuller test.

Due to the nature of the challenge, all teams approached it as a time series forecasting problem. Two teams modeled this problem using an Auto-Regressive Integrated Moving Average (ARIMA) algorithm - one team reported a mean squared error (MSE) of 0.3. One of those teams also used an Exponential Smoothing (ETS) algorithm and a Convolutional Neural Network (CNN), yielding much better results. Lastly, another team modeled the problem using a Gradient Boosting algorithm.

Main Insights from Data

During EDA, all teams found a significant discrepancy in the number of infractions by type - the following two being highlighted: street infractions and parking meter infractions. All teams also found that, on a road level, there are more parking infractions on arterial roads, such as West Broadway - mainly due to traffic intensity. Teams also found that the number of infractions has remained fairly stable throughout the years. 

As mentioned before, one team focused extensively on parking infraction causality derived from holidays and other events and plotted the chart in Figure 3. In red are the days with the least parking violations; in green are the days with the highest amount of parking violations, and in black are the bank holidays and major events in Vancouver.

Interestingly, it seems like Christmas Eve (December 24th) is systematically a day with a low number of parking infractions - it was marked as red and black in 2017. This could be caused by the smaller traffic flow as people are spending this day in particular mostly at home - however, no such hypothesis was proven using other sources of data. On the other hand, the team did go further into understanding the cause of the three peaks of traffic violations (marked green) by looking at big events on those days. They found that all of them had been days where major events took place in the city, such as concerts.

Figure 1 - Number of parking infractions (y-axis) per day (x-axis). The red, green, and black dots represent the local minima, maxima, and bank holidays, respectively.

Another team drew a map color-coded by the number of parking violations, as seen in  Figure 4. They found that most infractions occur in northern blocks, presumably due to the high concentration of services such as hotels, bars, restaurants, parks, shops, and an airport.

Figure 2 - Map of the parking infractions per street in Vancouver.

Lastly, one team also analyzed weather conditions against parking infractions and found that precipitation is an impact factor in general, as seen in Figure 3. Days with less rain correspond to more parking infractions - which can be expected, as that is usually associated with more people outside on the street and using their cars. When precipitation was above 50%, there were only 70 infractions per street daily, for example.

Figure 3 - Number of parking infractions (y-axis) compared to precipitation (x-axis).

Top 5 Solutions
Open-source code

Other challenges