Historic car park usage

Historic car park usage  

  By: jw35 on May 15, 2018, 4:03 p.m.

Our primary focus is on maintaining a real-time data processing platform, but we do maintain a historical record of the data we process. I'm currently working on a Web-based API that will make some of this historical data available for reuse. To give the current draft API a workout I've used our car park occupancy data to see if the removal of the £1/day charge for using the Park-and-Ride ar parks from 1 April this year had lead to an increase in usage.

Here's a quick-and-dirty graph of total parked-car-hours Monday-Friday for the Babraham Road car park:

Babraham Road usage

It's fairly clear that usage has gone up since 1 April. Much the same is visible in this graph for Trumpington:

Trumpington Road usage

though the effect isn't as marked. That could be because Trumpingdon in now routinely reaching capacity mid-week:

Trumpington capacity

Given that it also hit capacity in the runup to last Christmas even with a charge it looks as though this Christmas is going to be busy too.

The data is a bit noisy. Some of the downward spikes represent Bank Holidays when the usage is obviously reduced - I've already excluded Saturdays and Sundays becasue their lower usage just adds further noise. The other spikes represent problems with the received data which is sometimes missing, sometimes sticks at one value for a long time, and sometimes only grows throughout the day (suggesting a failure of the exit sensor). Graphs of the underlying data are available on the SmartCambridge web site (for example the current data for Babraham Road and Trumpington). The API will be available shortly.

Jon.

--
Jon Warbrick
Adaptive Cities Programme, University of Cambridge


 Last edited by: jw35 on May 21, 2018, 9:01 a.m., edited 1 time in total.

Re: Historic car park usage  

  By: edward on Aug. 3, 2018, 1:59 p.m.

This is very interesting, and important evidence with which to evaluate the decision to remove the £1 parking charge. See my questions in "Car park data problems" about cleaning the data. Here's a thought:

Add three extra data columns: 1) A quality flag: assign an R-squared value for fit to a normal distribution curve 2) Special status: bank holiday, closed for maintenance, special event, etc 3) Holiday status: school & university term time, school & university vacation, transitional period

This kind of analysis could also apply to traffic data, though with a different target distribution curve, and would make it easier to select meaningful and accurate subpopulation samples.

Edward

Edward Leigh Smarter Cambridge Transport