TransLink riders can better plan their journey on the bus network thanks to a successful pilot that implemented a new machine learning algorithm that improves accuracy of departure estimates.
The pilot program of this technology started in 13 buses and will now be affective system wide.
“We’re proud to have developed the new algorithm in-house, with collaboration from technology companies Microsoft and T4G,” said TransLink CEO Kevin Desmond. “This method is going to result in better information for customers who can make more informed decisions throughout their journey. During the pilot phase the difference between predicted and actual bus departure times improved by 74 percent.”
By combining live bus location data with the machine learning algorithm, this methodology improves existing estimates by considering major factors that affect bus departures. These include weather conditions and journey estimates at different times of day and night. To ensure accurate predictions for the entire transportation network, the algorithm involves over 16,000 machine learning models.
The new algorithm has been incorporated into TransLink’s Next Bus website and SMS tool (text the bus stop number and bus route number to 33333 for next departure estimates). Third-party applications that are already using our bus departure estimates, such as the Transit App and Google Maps, will also use the new estimation method.
