Google Maps 9.44 beta saw the debut of the new Parking Difficulty feature for some cities in the US, just a few weeks ago. In its recent post on Google Research Blog, the company has identified technical difficulties in calculating the parking availability using historical parking data for a particular time and day, weather, holiday/events, private parking lots and many such parameters.
The report adds that several issues stem from the false data sourced from private or gated parking spots, taxis dropping passengers at bus-stops and so on. The strategy employed to implement the Parking Difficulty feature was based on two factors: user feedback and Machine Learning.
The research team acquired user feedback data through questionnaires which revealed how long the user took to find a parking slot. The second method employed the use of Machine Learning from users who shared their location data to sum up info on parking difficulty, based on the user movement around the location of a parking spot.
Further explanation to the linear regression process comprising the parking difficulty model can be found on the Google Research Blog post that clearly explains the variables involved and false positives reported.
