Predictive Analytics: Unlocking New Avenues for Public Transportation

Anuj Khati
Srikar Manepalli
May 20, 2021 11 minute read

According to recent research from Valuates, the global smart transportation market is set to grow at a CAGR of 8.3% from 2021–2026, reaching a value of $121 million by the end of the forecast period.[1] One of the factors contributing to this growth is increasing government investment in smart cities and intelligent transport systems (ITSs), requiring cutting-edge AI & ML capabilities to process the vast volumes of data generated across transportation networks on a daily basis.

The footfall across public transportation networks has been steadily growing in the past few years. For instance, the number of unlinked passenger trips, i.e., the number of times passengers boarded public transportation vehicles, in the U.S., grew from 7.7 billion in 1995 to 9.95 billion in 2019,[2] signaling the rise of mobility as a cornerstone of the urban experience. However, the increase in ridership also brings about challenges of greater magnitude for public transportation players, including fluctuations in demand and extensive maintenance requirements.

Recent challenges created by the pandemic must also be considered at this point, which include steep declines in ridership rates and accompanying economic impacts. For instance, in 2020, New York City’s public transportation system reported a drop in subway ridership of 60% and bus ridership of 49% in comparison to 2019—contributing to a $12 billion fiscal deficit for public transport authorities.[3] Further, with inadequate capacity planning for buses and trains—for instance, with buses in Baku, Azerbaijan working on maximum capacity despite a 60% decrease in ridership during the pandemic[4]—larger structural flaws also became apparent. This, along with changes in people’s travel and commute patterns following the pandemic, are likely to have a massive impact on public transportation systems if not addressed effectively.

Newer concerns in the form of managing passenger traffic to tackle social distancing measures and crowding, as well as emerging mobility preferences, such as bike, car, and e-vehicle rentals, present a range of challenges for public transportation authorities to consider in the medium- and long term as well.

This is where the importance of data has become apparent in recent years—in terms of replacing time and resource-intensive methods of data collection, analysis, and reporting, and freeing information from various siloed databases to raise the level of interoperability. While analytics has been instrumental tackling traditional challenges, such as exponential passenger growth, it can also see implementation in anticipating fluctuations in demand, changes in demand and travel patterns, and pivoting in the aftermath of disruptions. Predictive analytical capabilities become imperative here in ensuring greater agility and operational efficiency, also doubling as a key starting point for digitalization and transformations in the public transport sector.

Enabling data-driven decision-making:

Predictive analysis can help public transportation authorities shift away from retroactive stock-taking to more forward-looking decision-making, going from asking questions such as ‘What happened?’ to more pertinent questions such as ‘What else can we do?’[5]

The utility of predictive capabilities becomes apparent in terms of planning new lines, routes, and services as well as pinpointing areas of cost and service reduction. It can be utilized at a business-wide level to explore avenues for growth and diversification. Business Intelligence (BI) dashboards that condense vast amounts of information across the transport network into customizable and intuitive visualizations can go a long way in facilitating decisions on procurement, services, customer outreach strategies, operational requirements, and infrastructure. Policymakers and executives can use such data to gain a real-time overview on multimodal transport networks and quickly implement decisions.

In addition to transforming raw data into meaningful output, predictive analysis can also form the basis of effective collaboration between various stakeholders in a public transportation setting—bringing an end-to-end understanding of the system, innovations based on expert opinions, and common communication frameworks within reach. Further, analytics can help ensure that each of these functionalities can be constantly improved upon, with regular, updated metrics on performance, quantifiable insights, and iterative processes to enable the continued competitiveness of decision-making tools.

Data derived from external sources, such as social media feeds, weather data, geospatial data, traffic monitoring mechanisms, parking services, and so on, can further enrich business decisions by providing foresight on the transportation system as a whole. This becomes key to developing holistic solutions for commuters.

Predictive capabilities can also be used at the ground level to optimize operations including rail switching, ticketing, infrastructure development projects, staffing, bus deployment, scheduling, and so on. For instance, in Los Angeles, an electric scooter-based service was linked to a city-operated database, with each commuter’s journey mapped, analyzed, and used to understand traffic flow patterns. Such applications can be pivotal to determining optimal traffic conditions and easing urban congestion in the long term.[6]

In addition to facilitating streamlined customer-facing services, predictive capabilities become a crucial part of the transportation system itself, helping address key concerns for public transportation players, such as maintenance and route optimization.

Optimizing predictive maintenance:

Predictive tools and AI-driven fault identification systems can greatly simplify maintenance processes across transportation networks, including routine checks, asset ageing checks, asset replacement, and power supply maintenance, through connectivity-enabled infrastructure. Data from across sources, including historical maintenance records, telematic sensors, and human input, can be analyzed through a combination of predictive algorithms and ML to create a self-learning and intelligent prediction model, which can help prioritize repair and replacement requirements.

For example, 250 locomotives from a Europe-based freight carrier were retrofitted with performance management software to examine brake performance, motor temperature, and other conditions to predict maintenance. The system, which fed on real-time data rather than pre-defined metrics, proved to have greater accuracy and reduced failure rates by 25%.

Moreover, condition-based and predictive maintenance can significantly reduce maintenance costs in addition to reducing down time and system lags. ML-based predictive capabilities can further indicate vehicles to be taken out of circulation and even predict the chances of employee absences through weather, road, and traffic analysis, ensuring that the risks of service disruptions are accounted for and minimized.

Improving route optimization and ease of travel:

Data on the real-time location of trains, buses, and so on can be used to send out automated alerts to passengers in the case of disruptions and delays as well as give them updated information on arrival times. Predictive analyses can play a significant role in indicating the possibilities of weather impacts, accidents, vehicle breakdowns, maintenance projects, and strikes, among others, by studying data sourced from route segment levels. Following this, alternate routes and services can be charted out using a combination of GPS functions, traffic monitoring systems, and automated applications, and in addition, more vehicles can be deployed, staffing increased, and advance planning undertaken to ensure a seamless experience. For example, the city of Pittsburgh implemented an AI-powered traffic light system that factored in real-time conditions instead of following pre-set patterns. A predictive model was employed to inform traffic signaling patterns, and the data gathered from the system helped reduce vehicles’ braking by 30%, travel time by 25%, and engine idling by more than 40%.[7]

Passenger focus: Convenience and security.

As ridership numbers across all forms of public transport have risen significantly, several public transportation players have turned their focus to convenience to improve overall customer service. From providing information on various transport options, including taxis, cycles, and trains, developing journey planning applications, offering variety in terms of connecting routes, offering sustainable transportation options such as e-bikes and e-scooters, touchless payment options, and personalized notifications based on location tracking, the applications of AI have ranged far and wide.

Predictive analyses play their part here in improving the customer-centric experience: indicating where passengers are likely to disembark, offering information on train/bus capacity,[8] and predicting peak times to help passengers avoid crowds. Information on likely road conditions, weather, and vehicle movement becomes readily available to end users, and interactive maps that display real-time information sourced from buses, trains, and other forms of transport can offer convenience to both transport authorities and customers alike. Applications such as these become key to reducing the unpredictability of public transport services, which continues to be a major deterrent for commuters across the globe.

Safety is another important factor that can be improved through predictive analyses. For instance, several railway networks have installed facial recognition cameras across stations to detect disruptive or dangerous behavior. With AI able to analyze facial expressions, body language, and so on—while keeping data anonymous—the ability to proactively identify threats becomes crucial for transport authorities to act on. Predictive analytics also comes into play in order to reduce accidents on the road. For instance, the state of Tennessee launched a crash prediction program in 2013, examining traffic citations, crash reports, weather patterns, and events using AI to generate crash-prediction maps. This move led to a 3% drop in traffic fatalities between 2013 and 2015, and reduced accident response time from 28 minutes to 22 minutes.[9]

Automating demand prediction and pricing:

Predicting peak periods, including those related to holidays, cultural and sporting events, and famous attractions, is a significant factor for public transport authorities to consider when planning operations, services, and schedules. The ability to foresee such peaks can provide a competitive advantage in terms of adequately meeting demand, reducing operating costs, and improving service quality. Predictive modeling can also positively impact the greater move towards personalization in the public transport sector, in terms of being able to gauge customer requirements and send personalized reminders and notifications. For example, Dubai Metro introduced an AI-based system to indicate optimal travel times for individuals, thereby reducing crowding and adhering to distancing norms.[10]

Pricing and payment methods become another differentiator for public transport authorities to consider, especially to achieve customer retention and encourage customer loyalty. Predictions based on usage patterns can be leveraged to offer custom payment and subscription plans, discounts, passes, and so on, ensuring greater customer engagement and making flexible options readily available. For example, the half-fare travelcard offered by Switzerland reduces all ticket costs by 50% and is considered one of the most popular travel subscription services in the country. Further discounted rates are offered to people aged 16–25 to encourage public transportation use, additional services and flexible mobility options are used to incentivize commuters, and loyalty prices are also offered to encourage repeat customers.[11] With the rise of Mobility as a Service (MaaS) across the globe, pay-as-you-go options and mobility packages are also seeing increased adoption. For instance, in Quebec, municipal transport authorities are offering mobility packages that include bike sharing and carsharing, with a commuter being able to save on public transport costs by purchasing a single, cost-effective package.[12] Customer requirements such as these can be accounted for to create more attractive pricing plans, therefore enabling sustained engagement.

As public transportation authorities are set to deal with an increasing number of travelers year on year, with urban centers continuing to expand and data generated across networks multiplying exponentially, predictive capabilities will be key to ensuring agility, competitiveness, and the ability to adapt to a rapidly changing market.



Anuj Khati
Senior Associate

Srikar Manepalli
Senior Associate