Abstract
Official road classification is used for general purposes but for deep traffic analysis this classification is not sufficient. Today there are efficient ways to collect large amounts of data from multiple sources that can be used for different causes. These large amounts of data cannot be analysed with traditional methods and new state-of-the-art algorithms should be used. The paper presents the methodology for urban road classification based on GPS (Global Positioning System) vehicle tracks and data on infrastructural characteristics of road subsegments. The process of defining road categories includes data collection and analysis, data cleansing and fusion, multiple regression, principal component analysis (PCA) as well as cross-validation and k-nearest neighbour (kNN) classification procedure. Results of such continuum can be used as base for further traffic analysis as travel time prediction, optimal route detection etc.References
