Travel Time Estimation Results with Supervised Non-parametric Machine Learning Algorithms

摘要

Paper describes urban travel time estimation procedure based on non-parametric machine learning algorithms and three data sources (GPS vehicle tracks, meteorological data and road infrastructure data base). After data fusion and dimensionality reduction, new road classification is defined and for four different time intervals and five different road categories travel time estimation is conducted. For travel time estimation, k nearest neighbors (kNN) and IRM-based (Iterative Regression Method) approaches were applied. Best results for two hour forecasting period are achieved for road class with highest traffic flow.