As is the case with many technology-related market sectors today, next-generation transportation analytics is evolving with tremendous pace. Newer transportation analytics are transitioning from first-generation “descriptive analytics”, which utilizes historical analysis of the transportation environment, to the use of “predictive” applications, which aggregate greater data sets and apply enhanced algorithms to generate forecasts for transportation systems. The latest generation of transportation analytics now incorporates “prescriptive” analytics, where real-time, unique user-based solutions are developed.
Recently we’ve seen the emergence of predictive analytics take root in the transportation industry. To date, predictive analytics have predominantly focused on forecasting congestion patterns in support of trip and route planning, and utilized for the derivation of future travel time projections. Predictive analytics move well beyond descriptive analytics by including the aggregation of larger, disparate data-sets including real-time and historical data, and data processing that includes laws of probability, statistical modeling, game theory and advanced analytic algorithms in order to generate estimated future conditions.
Predictive analytics are now including the human element, through the use of “social” algorithms such as “collaborative filtering”. This process aggregates large, diverse data sets and applies advanced predictive analytics in order to generate tailored, unique end-user information. This form of real-time predictive analytics model includes the integration of a wide range of user-based data in order to determine individual preferences, or unique user profiles. Collaborative filtering was first implemented in the internet community by such companies as eBay, Netflix and Amazon to predict user preferences and potential interest areas. The best known example of collaborative filtering can be found in algorithms developed for the Netflix Prize.
Collaborative filtering techniques will become a valuable real-time analytic tool for the transportation community, as it begins to better define unique user-needs and gain the ability to generate higher resolution user profiles. The wide-scale proliferation of data sources, including smartphone, vehicle and RFID, has greatly enhanced the data landscape in the transportation environment. Algorithms will learn user preferences, trends, needs and constraints, build qualitative user profiles that are fused with real-time data and predictive analytics to generate tailored, individual traveler solutions.
Prescriptive analytics represent the leading edge of transportation analytics. Prescriptive analytics take predictive analytics to the next level by generating solutions through the fusion of real-time situational awareness, unique individual end-user preferences and user profiles, with advanced analytics to generate solutions and recommendations on a user by user basis. The combination of predictive and prescriptive analytics represent extremely powerful tools, more powerful than human capabilities, because of the wide range and volume of data and information aggregated and processed, and the advanced algorithms applied to the aggregated data sets. Predictive and prescriptive analytics will also assist in removing the burden of “data overload” from the traveling public, by understanding unique user needs, generating individual traveler profiles and generating best-case transportation solutions on an individual user basis.
Next-gen transportation analytics will continue to model and forecast conditions for transportation systems. Over time, these tools will provide a significant resource to the transportation industry. New prescriptive analytics will provide valuable tools for transportation operations entities, as well as for the transportation end-user (traveler). However, we are just moving beyond the first generation (descriptive) analytics and into the predictive analytics arena. To date, predictive analytics have shown mixed results during their initial rollout. For example, Google’s predictive travel time service was recently shut down, apparently due to inconsistencies and reliability issues. Over time, these tools will only get better, but for now, the debate will continue regarding decision-grade predictive and prescriptive analytics and minimum reliability thresholds.
IBM Brings Watson-Like Sleuthing to Solve Traffic Crisis
Predictive Traffic More than Good Enough
Descriptive, Predictive, & Prescriptive analytics