Applying K-Nearest Neighbor Algorithm for Statewide Annual Average Daily Traffic Estimates

Assigning non-ATR sample count sites to different factor groups is an imprecise process. Currently, factor groups are determined on the basis of a combination of geographic location and functional roadway classification. This paper proposes a new K-nearest neighbor algorithm using geographic information system (GIS) technology. Roadway and land use characteristics can be captured in the K-nearest neighbor algorithm for the factor group process. The simulation results show that an unweighted K-nearest neighbor algorithm can produce better AADT estimates than the traditional eighty-four factor approach that uses each functional class as a factor group. The K-nearest neighbor algorithm can be a useful way to carry out roadway classification.