{"id":170817,"date":"2011-09-26T11:31:07","date_gmt":"2011-09-26T11:31:07","guid":{"rendered":"https:\/\/www.microsoft.com\/en-us\/research\/project\/predictive-analytics-for-traffic\/"},"modified":"2019-08-19T15:07:23","modified_gmt":"2019-08-19T22:07:23","slug":"predictive-analytics-for-traffic","status":"publish","type":"msr-project","link":"https:\/\/www.microsoft.com\/en-us\/research\/project\/predictive-analytics-for-traffic\/","title":{"rendered":"Predictive Analytics for Traffic"},"content":{"rendered":"

Machine Learning and Intelligence for Sensing, Inferring, and Forecasting Traffic Flows<\/h2>\n

Machine learning and intelligence are being applied in multiple ways to addressing difficult challenges in multiple fields, including transportation, energy, and healthcare. Research scientists at Microsoft Research have been engaged in efforts in all of these areas. We focus on multiyear efforts at Microsoft Research to infer and forecast the flows of traffic. The work leverages machine learning to build services that make use of both live streams of sensed information and large amounts of heterogeneous historical data. This has led to multiple prototypes and real-world services such as traffic-sensitive directions in Bing Maps. Focused work in this realm also stimulated new efforts in related areas, such as privacy and routing.<\/p>\n

About<\/h2>\n
\"Clearflow

Fielded Smartphlow traffic application.<\/p><\/div>\n

Traffic has been growing in major cities around the world given the increase in densities of cars on roads and the slow development of road infrastructure. With research starting in 2002, research scientist and developer teams at Microsoft Research pioneered the use of machine learning methods to build predictive models for traffic. The work led early on to prototypes that can infer and predict the flow of traffic at different times into the future based on the analysis of large amounts of data on traffic over months and years. The work was leveraged in revolutionary services, such as traffic maps that show users how traffic is evolving over time, as well as in services that provide traffic-sensitive directions by considering the inferred speeds on roads that are not sensed directly.<\/p>\n

Research on machine learning for traffic spanned several projects and has focused on both on principles and applications. Multiple technical and empirical studies were performed as part of this work. On the fielding of applications, the research efforts sit behind the traffic-sensitive directions in Bing Directions (opens in new tab)<\/span><\/a> within Bing Maps and the mobile directions service on the Windows Mango phones. A portion of Microsoft Research\u2019s methods, tools, and software on predictive analytics for traffic were licensed externally in 2004 to traffic startup Inrix (opens in new tab)<\/span><\/a> shortly after the company was formed, helping to slingshot that company into the world as a leading international provider of traffic analyses and predictions.<\/p>\n

As part of efforts on learning about traffic flows from data, researchers at Microsoft Research explored methods that enhance the safety and privacy of people who wish to help with the \u201ccrowdsourcing\u201d of real-time flows of road data from their mobile GPS data. Principles of community sensing<\/i> have been developed. These principles center on working with people under a \u201cprivacy budget\u201d based on the use of the computations of the value of information for understanding flows over time on the road network.<\/p>\n\t\t\t

\n\t\t\t\t\t\t\t
\n\t\t\t\t\t\n\t\t\t\t\t\t