Yarra Trams operates Melbourne’s tram network, the world’s largest spanning across 250 kms of double track, consisting of more than 500 trams across 7 classes. This 24/7 network enables more than 5000 tram services daily, and more than 200 million passenger trips completed each year. Portable has been working with Yarra Trams since 2013 to optimise network management systems and provide ongoing software support to ensure the team are equipped with bespoke and configurable technical solutions, to continue meeting the demands of this busy public network.
Manual Attribution Tool
One of these technologies is the Manual Attribution Tool (MAT), which is a back-end system designed for Operations Centre staff to meet daily incident reporting requirements to the Victorian Government (through Public Transport Victoria) and internal Yarra Trams reporting requirements.
Portable worked with Yarra Trams to deliver a powerful tool that enables Yarra Trams to investigate and understand why trips are cancelled or delayed, and measures trends and insights which informs how Operations Centre staff mitigate potential incidents and inform future operational planning and logistics.
Operations Centre staff have access to an intuitive web interface that loads real-time data and metrics, including incidents reported by drivers out on the network, enabling them to quickly filter the current day's trips, and attribute them to the incident.
Of the 200 million passenger trips completed each year, delays occasionally occur, mostly due to incidents such as traffic obstructions, bursts of increased passenger traffic, anti-social commuter behaviour, technical malfunctions, and several unpredicted causes. An opportunity arose for Portable to assist with using machine learning to automate some of the analysis of understanding why delays occur. This MAT Suggestions Tool is a new feature to the MAT system. It uses predictive analytics and automated decision making to reduce the manual work in the Yarra Trams Operations Centre and improve accuracy in the management of tram trip data.
Design challenge
The Yarra Trams Operations Centre are responsible for essential tasks such as incident management, operational planning, driver communications and passenger needs, and are a critical part of keeping the network running smoothly and seamlessly.
Management of tram data and understanding why a tram trip may have been interrupted or delayed is time consuming for Yarra Trams. This work has historically relied on many time consuming complex steps, looking at data across multiple systems and making a human judgement call sometimes leading to inconsistencies. Any recurrent process with a high volume of data is an opportunity for the use of artificial intelligence - for machine learning software to automate repetitive tasks for higher levels of accuracy and in a fraction of the time.
Yarra Trams and Portable set out to develop and test a concept of how using predictive analytics could improve the efficiency and accuracy of understanding of tram trip interruptions or delays.
Approach
Gathering requirements
Portable and Yarra Trams mapped out all the data points that the Yarra Trams Operations Centre use for the analysis of trip interruptions and delays including:
- Data from tram drivers,
- Reported events from Victoria Police, VicRoads and other transit and government authorities of interruptions to the network
- Tram timetables
- Real time tram data
- Trip journey time across certain sections of the route
We engaged in a co-design activity with Yarra Trams Operation staff to understand current repetitive manual processes of using these data sources and how we could use real time data integration and machine learning to automate these processes and improve accuracy.
Identifying the goals of the solution
Together we defined the problem statement the solution needed to address. We understood the goal to develop a powerful tool that combines all of these data points and enables Yarra Trams to investigate and understand why trips are cancelled or delayed. The tool would also allow Yarra Trams to measure trends and insights which informs how Operations Centre staff mitigate potential incidents and inform future operational planning and logistics.
Utilising predictive analytics
Additionally, we worked with Yarra Trams to design a data concept that found patterns in day to day system interactions. Mapping these patterns we were then able to implement predictive analytics to design a feature that suggests interrupted trips. This used configurable criteria such as time delays across part of a route and data points from a range of sources to present to the Operations Centre team members possible events that have caused delay.
Proof of concept
Using an agile process we developed a proof of concept for validation by the Yarra Trams Operations Centre team. This proof of concept used machine learning to analyse timetable data, real time tram trip data, data The tool uses this data to provide a system prompt to suggest likely delays and the likely cause of them.
Pilot and testing
In the first stage, each system prompt was reviewed by a Yarra Trams Operations Centre team member to review the accuracy. We piloted the new tool, keeping a close eye on accuracy and staff feedback and reviewing how the tool was learning. Even in this first stage using machine learning moved the staff effort from completing manual data management steps to just making a judgement call based.
Managing continual improvements
We built a continual improvement framework where we worked with Yarra Trams to continually review and improve the tool. In the second stage of the project, we then began to make further usability enhancements based on feedback from the team at Yarra Trams using the service and automate more of the manual steps involved in managing tram interruptions.
Outcomes
This was an exciting project for Portable and Yarra Trams to solve a real world use case using new advances in data technology.
- The tool has minimised the manual effort of the Operations Centre staff, and automated the incident reporting system to enable more efficient and effective reporting and insights.
- The tool identified trips that would otherwise have been missed by the Operations Centre staff.
- It has automated the analysis of over 2000 trips a month reducing the amount of manual data analysis work by Yarra Trams Operational Staff.
- In its first 3 months through as staff ongoing review and quality audit the tool has shown a 80% accuracy rate
Project team
Anthony Daff, Lead Developer
Lincoln Le, Senior Developer
Jimit Rajani, .NET Developer
Chris Rickard, Lead Developer
Caitlin Williams, Senior Producer
Naomi Wilson, Senior Client Partner
Sign up to our email newsletter to get updates about our events, work and research
