Machine-learning program predicts public transport use in Singapore 

January 25, 2017
Machine-learning program predicts public transport use in Singapore 
From left: Gary Lee, Hu Nan, Erika Fille Legara, and Christopher Monterola. Credit: A*STAR Institute of High Performance Computing

From schools and shops to hospitals and hotels, a modern city is made of many different parts. Urban planners must take account of where these services are located when designing efficient transit networks. A*STAR researchers have developed a machine-learning program to accurately recreate and predict public transport use, or 'ridership', based on the distribution of land-use and amenities in Singapore.

Traditional cities comprise an inner central business district (CBD), where most people work, surrounded by outer residential and industrial zones. Unfortunately for commuters, the high volume of people traveling to and from the CBD can cause gridlock at peak hours. To alleviate some of this frustration, the Singaporean government is working on creating regional centers by the year 2030. The planners hope to encourage business owners to open at specified regional centers around the city-state, easing peak-time pressure and encouraging use.

"We're aiming to understand the recipe for a smart city," explains Christopher Monterola at the A*STAR Institute of High Performance Computing, who led the project in collaboration with scientists across Singapore. "Singapore needs an efficient transport system to support people's activities given the existing and planned infrastructure. To guide planners, we needed a model that could predict ridership under the regional centers plan."

The team collected data from the city's smartcard system on people tapping in and out of individual bus and subway stations over a period of a week—more than 20 million journeys in total.

The smartcard data was combined with city-wide information on how land was being used—for business, industry, residence, water or greenery—and high-resolution maps that identified individual amenities within a set radius of each station. Monterola's team trialed three different machine-learning models—computer programs that train themselves through repeated simulations—to find one that first accurately reproduced, and then predicted, transport ridership across the city.

"We found that a decision tree model performed best, with good accuracy, computational efficiency and an easy-to-follow user display," says Monterola. "Results indicated that an increase in amenities of up to 55 per cent across the city would increase ridership. Beyond this point, ridership begins to decline; this is logical because if amenities are available locally, people walk instead."

The high-resolution amenity data proved a much stronger predictor of ridership than general land-use details; a useful result for informing future urban planning and monitoring Singapore's regional centers as they develop. The model could be applied to any city with access to similar high-resolution data, notes Monterola.

Explore further: New computer model of city dynamics could pave way to planning sustainable urban areas

More information: Nan Hu et al. Impacts of land use and amenities on public transport use, urban planning and design, Land Use Policy (2016). DOI: 10.1016/j.landusepol.2016.06.004

Related Stories

An agent-based model that analyzes commuters' travel data

November 19, 2014

The introduction of smartcard ticketing for Singapore's public transport system has enabled A*STAR researchers to provide valuable predictive data on potential train overloading.  This will enable system planners to address ...

Recommended for you

Lightning sparking more boreal forest fires

June 27, 2017

A new NASA-funded study finds that lightning storms were the main driver of recent massive fire years in Alaska and northern Canada, and that these storms are likely to move farther north with climate warming, potentially ...

Collapse of the European ice sheet caused chaos

June 27, 2017

Scientists have reconstructed in detail the collapse of the Eurasian ice sheet at the end of the last ice age. The big melt wreaked havoc across the European continent, driving home the original Brexit 10,000 years ago.

Greenland now a major driver of rising seas: study

June 26, 2017

Ocean levels rose 50 percent faster in 2014 than in 1993, with meltwater from the Greenland ice sheet now supplying 25 percent of total sea level increase compared with just five percent 20 years earlier, researchers reported ...


Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.