Machine learning for ride-hailing service, Grab
EITN speaks with Head of Engineering at Grab, Ditesh Gathani, about machine learning, and how it’s raising the bar for the ride-hailing industry.
EITN: What is the top use case for machine learning within Grab?
Ditesh: The Grab machine learning team works on all kinds of predictions using traditional machine learning and new deep learning techniques. Most applications on Grab involve studying our users’ behaviour to improve the experiences for both our passengers and driver-partners. For example – it is thanks to Machine Learning that the app auto-detects a passenger’s pickup and drop-off location preferences.
The app will automatically trace the riders’ everyday routine at a specific time and place so that the users don’t have to fill-in the details of pickup and drop-off location. Similarly, our telematics programme that leverages on the combination of machine learning and predictive analytics to identify unsafe driving habits by drivers such as speeding, swerving and hard braking or accelerating, and educating drivers about ways to improve the driving experience. Drivers will then receive a report every week for their reference and further action. And thanks to the telematics initiative, traffic accident rates in Malaysia with Grab are on average 6 times lower than the country average.
EITN: What are the top reasons for use of technology within Grab? What are the top objectives you try to achieve with technology?
Ditesh: As our co-founder Anthony Tan put it, through the power of technology Grab aims to become an active contributor to solving Southeast Asia’s fundamental issues: congestion, jobs, trust in and access to the digital economy. We leverage technology to disrupt the norm whereby we find better solutions to do things such as hailing a ride, enhancing safety, improving welfare and improving the cities we are in. More specifically, some of the way we are harnessing tech to improve the lives of our passengers and drivers, are driver preferences and utilisation.
We’ve optimised our driver booking system by combining the power of our localised data-set combined with machine learning. We have learned our drivers’ preferences and behaviours, enabling us to predict which jobs drivers will take and assign targeted jobs. Bookings are then sent to drivers with the highest probable booking rate. This helps drivers be more engaged by giving jobs they like. We have just rolled out a new feature for our drivers under the Better 365 programme. Though in beta mode, My Destination is a feature for our drivers that allows them to select their final destination and the app will send them jobs along the way
Our data science team analyses real-time and historical demand data to alert drivers to anticipated demand hotspots. Our model is sensitive enough to account for large range of factors, including the arrival of ferries, weather and days of the week. From this, we have managed , to further enhance our driver-partners experience, whereby SMSes are sent to drivers to notify them of high-demand areas. For example, a flight or a outstation bus just arrived at their stop. This will enable them to be more efficient in serving passenger demand and earn more in the process. Based on this efforts, our internal study conducted in June 2017 shows Grab drivers earn on average one-third (32%) more per hour than the average worker in their country. In Malaysia that number is slightly higher at 48%.
Besides that, we also use tech to help maximise the time for our drivers on the road while improving their income and we do so by leveraging on the data that we have.
For our passengers, we have rolled out safety initiatives such as Grab’s telematics programme as upgrades to the app which make the overall user experience a pleasant one. One example of a seamless user experience is the predictive pick up and drop up locations – with machine learning, the app learns a passengers favourite location and pre-fills the POIs.
Overall, our top objective is to use tech to improve our services and features for both our driver-partners and our passengers which include enhancing safety, improving the efficiency of our services for our drivers on the road.
EITN: For machine learning to work, I presume there would have to be a huge collection of data to perform analytics upon. If true, how do you ensure that privacy of your user base, is protected and not abused?
Ditesh: User Trust and Safety is paramount to Grab – we uphold the industry’s highest security standards on our GrabPay platform. A large part of what underlies this protection is the risk and fraud detection system that we put in place at the launch of GrabPay early this year. It continues to evolve further today, using sophisticated machine learning algorithms that progressively builds on the knowledge we have of our drivers, passengers and their travel patterns to enable the largest mobile transaction volume on any Southeast Asian consumer platform in history.
Grab has always placed security and privacy as their top two priorities in order to avert any possible threats that would undermine system. Internally, Grab has created a term in their operations called “3S” which stands for Security, Stability and Scalability. Security is always on top as it is important to ensure everything is secure; stability, is how Grab make sure the platforms are reliable, and scalability is when we need to scale our system in case there is a sudden demand due to a certain reason.
Furthermore, to keep the data of Grab riders and passengers secure, a multi-layered approach is implemented where the data and access to the infrastructure are tightly controlled by particular people that can have access to it.
Another internal security approach that takes place is the availability of a data privacy officer who ensures that the right processes are done and all the government regulations are followed.
EITN: How do you go from 10TB of data, daily towards helping governments achieve smart cities?
Ditesh: In Malaysia, Grab has collaborated with Multimedia Development Corporation (MDeC) to address traffic congestion and road safety challenges in major Malaysian cities. We have made improvements in transport infrastructure and traffic management that will bring much needed relief to the increasing number of motorists in Malaysian urban hotspots who are braving the daily traffic gridlock during their commute.
According to Nielsen, Malaysia has the 3rd highest car ownership in the world. Furthermore, a 2015 World Bank report found that Kuala Lumpur’s population alone wastes up to 500 million hours of work a year, idling in traffic and burning up to 1.2 billion litres (about 315 million gallons) of fuel. In sum, these losses are estimated to exceed 2 percent of the country’s gross domestic product.
To combat road congestion, the government has expanded the public transit network and infrastructure. However, not all LRT / MRT stations are easily accessible and many Malaysians still have to drive or rely on other forms of public transport to get to the stations. This is where Grab hopes to fill in the gap for first mile and last mile travel by providing safe and reliable transport.
Grab’s main objective is to accommodate the mobility needs of Malaysians during peak-hour traffic. During peak hours, demand for point-to-point transportation can increase as much as three times – a huge pain-point for Malaysians.
Leveraging on our technology, Grab has been able to effectively match the supply and demand for vehicles in priority locations.
Additionally, Grab also provides our driver location data to the OpenTraffic platform, a collaboration with the World Bank to provide Southeast Asian governments (MY, ID, PH) access to real-time traffic information.