Supervised Machine Learning Applications to Winter Road Impacts
Abstract
Winter weather causes major disruptions to road travel, increasing delays and accident rates. Significant investments are made towards maintaining the condition and safety of the road network. Weather forecasts provide the state of the art prognosis for winter weather phenomena, but are not designed to explicitly predict road surface conditions. This thesis implements supervised machine learning techniques, most successfully random forests, using weather forecast data to directly predict the Winter Driving Index of roadways as defined by the Indiana Department of Transportation (INDOT). Attempts are made to also predict the amount of traffic reductions due to winter weather using random forests and artificial neural networks. Different case studies are presented to demonstrate successes and pitfalls of the models, and possibilities for future development are discussed.
Degree
M.S.
Advisors
Baldwin, Purdue University.
Subject Area
Atmospheric sciences
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.