Automated Machine Learning Pipeline for Traffic Count Prediction

Alireza Shojaei,Haluk Laman,Amr Oloufa,Milad Salem,Jiann-Shiun Yuan,Amirsaman Mahdavian

doi:10.3390/modelling2040026

Abstract

Research indicates that the projection of traffic volumes is a valuable tool for traffic management. However, few studies have examined the application of a universal automated framework for car traffic volume prediction. Within this limited literature, studies using broad data sets and inclusive predictors have been inadequate; such works have not incorporated a comprehensive set of linear and nonlinear algorithms utilizing a robust cross-validation approach. The proposed model pipeline introduced in this study automatically identifies the most appropriate feature-selection method and modeling approach to reduce the mean absolute percentage error. We utilized hyperparameter optimization to generate a universal automated framework, distinct from model optimization techniques that rely on a single case study. The resulting model can be independently customized to any respective project. Automating much of this process minimizes the work and expertise required for traffic count forecasting. To test the applicability of our models, we used Florida historical traffic data from between 2001 and 2017. The results confirmed that nonlinear models outperformed linear models in predicting passenger vehicles’ monthly traffic volumes in this specific case study. By employing the framework developed in this study, transportation planners could identify the critical links on US roads that incur overcapacity issues.

Highlights

Growth in the number of vehicles and degree of urbanization mean that the annual cost of traffic jams is increasing in cities
Fu and Kelly [1] employed Artificial Neural Network (ANN) versus log-linear and ordinary least squares (OLS) approaches to predict traffic volume. Their comparison results showed that the ANN method achieved a mean absolute percentage error (MAPE) of 28.58%, which meant that it outperformed the log-linear model (52.49% MAPE) and OLS (66.6% MAPE)
The random forest (RF), K-Nearest Neighbor (KNN), Decision tree (DT), and ANN models performed the best when trained on the training data set and tested on the validation data set

Summary

Introduction

Growth in the number of vehicles and degree of urbanization mean that the annual cost of traffic jams is increasing in cities. This leads to a decreased in the quality of life among citizens through a considerable waste of time and excessive fuel consumption and air pollution in congested areas [1]. Research on predicting the traffic data is essential. Such studies are valuable for planning the allocation of limited resources to highways that are most at risk for experiencing congestion and for developing an improved intelligent traffic management service [1]. The selection of a suitable algorithm to project traffic volumes is essential

Objectives

Methods

Results

Discussion

Conclusion