Predictive modeling of critical headway based on machine learning techniques

Abstract

Due to the impossibility of directly measuring of critical headway, numerous methods and procedures have been developed for its estimation. This paper uses the maximum likelihood method for estimating the same at five roundabouts, and based on the obtained results and pairs of accepted and maximum rejected headways, several predictive models based on machine learning techniques were trained and tested. Therefore, the main goal of the research is to create a model for the prediction (classification) of the critical headway, which as inputs, i.e. independent variables use pairs - accepted and maximum rejected headways. The basic task of the model is to associate one of the previously estimated values of the critical headway with a given input pair of headways. The final predictive model is chosen from several offered alternatives based on the accuracy of the prediction. The results of training and testing of various models based on machine learning techniques in IBM SPSS Modeler software indicate that the highest prediction accuracy is shown by the C5 decision tree model (73.266%), which was trained and tested on an extended data set obtained by augmentation or data set augmentation (Data Augmentation - DA).

Keywords

• critical headway
• predictive modeling
• machine learning
• da
• decision tree