MACHINE LEARNING SOLUTIONS FOR TRANSPORTATION NETWORKS : Learning the behavior of traffic flow

Tomas Singliar

ISBN 10: 3639171608 / ISBN 13: 9783639171600

Editore: VDM Verlag Dr. Müller, 2009

Riguardo questo articolo

nach der Bestellung gedruckt Neuware - Printed after ordering - This thesis brings a collection of novel models andmethods that result from a new look at practicalproblems in transportation through the prism of newlyavailable sensor data. From this data, we build a model of traffic flowinspired by macroscopic flow models. Unliketraditional such models, our model deals withuncertainty of measurement and unobservability ofcertain important quantities and incorporateson-the-fly observations more easily. Having apredictive distribution of traffic state enables theapplication of powerful decision-making machinery tothe traffic domain.Secondly, a new method for detecting accidents andother adverse events is described. Data collectedfrom highways enables us to bring supervised learningapproaches to incident detection. However, a majorhurdle to performance of supervised learners is thequality of data which contains systematic biasesvarying from site to site. We build a dynamicBayesian network framework that learns and rectifiesthese biases, leading to improved superviseddetector performance with little need for manuallytagged data. The realignment method applies generallyto virtually all forms of labeled sequential data.

Codice articolo 9783639171600

Dati bibliografici

• Titolo: MACHINE LEARNING SOLUTIONS FOR TRANSPORTATION NETWORKS : Learning the behavior of traffic flow
• Editore: VDM Verlag Dr. Müller
• Data di pubblicazione: 2009
• Lingua: Inglese
• ISBN 10: 3639171608
• ISBN 13: 9783639171600
• Rilegatura: Taschenbuch
• Condizione: Neu
• Dimensioni: 220 L x 150 A x 8 P (millimetri) 220 millimetri di larghezza per 150 millimetri di altezza per 8 millimetri di profondità
• Peso: 188 grammi

Descrizione articolo

Riassunto

This thesis brings a collection of novel models and methods that result from a new look at practical problems in transportation through the prism of newly available sensor data. From this data, we build a model of traffic flow inspired by macroscopic flow models. Unlike traditional such models, our model deals with uncertainty of measurement and unobservability of certain important quantities and incorporates on-the-fly observations more easily. Having a predictive distribution of traffic state enables the application of powerful decision-making machinery to the traffic domain. Secondly, a new method for detecting accidents and other adverse events is described. Data collected from highways enables us to bring supervised learning approaches to incident detection. However, a major hurdle to performance of supervised learners is the quality of data which contains systematic biases varying from site to site. We build a dynamic Bayesian network framework that learns and rectifies these biases, leading to improved supervised detector performance with little need for manually tagged data. The realignment method applies generally to virtually all forms of labeled sequential data.

L'autore

Tomas specializes in machine learning and anomaly detection, especially by means of graphical probability models. He obtained his PhD from University of Pittsburgh in 2008, authored papers on inference in graphical models, modeling of large sensor data sets and anomaly detection and served on program committees of several major AI conferences.