Night Traffic Flow Prediction Using K-Nearest Neighbors Algorithm

Abstract

The aim of this research is to predict the total and average monthly night traffic on state roads in Serbia, using the technique of supervised machine learning. A set of data on total and average monthly night traffic has been used for training and testing of predictive models. The data set was obtained by counting the traffic on the roads in Serbia, in the period from 2011 to 2020. Various classification and regression prediction models have been tested using the Weka software tool on the available data set and the models based on the K-Nearest Neighbors algorithm, as well as models based on regression trees, have shown the best results. Furthermore, the best model has been chosen by comparing the performances of models. According to all the mentioned criteria, the model based on the K-Nearest Neighbors algorithm has shown the best results. Using this model, the prediction of the total and average nightly traffic per month for the following year at the selected traffic counting locations has been made.

Keywords

• machine learning, traffic flow, prediction, K-Nearest Neighbors, Weka