Application of artificial intelligence models in traffic flow prediction and time-of-day breakpoints determination

Abstract

This research aims to solve the problem of intersection traffic flow prediction along with the determination of time-of-day (TOD) break points for a pre-timed or actuated intersection. This study investigated mainly three artificial intelligence based models namely group method of data handling (GMDH) model, adaptive neuro-fuzzy inference system (ANFIS) model and type-2 fuzzy logic (FL) model to predict intersection traffic flow considering spatial and temporal characteristics. It is observed that that fuzzy c-means (FCM) clustering algorithm based ANFIS models and type-2 FL models outperform other considered models for two approaches of the considered intersection. Based on the obtained performance measures, it can be concluded that all the considered models are valid and promising for predicting traffic flow. This study also proposes a novel methodology in which time variable and Z-score of the approach traffic counts are used as the prospective features for determining TOD breakpoints instead of relying on the judgmental approach. The obtained results solve the problem of frequent changes of TODs. At the end, this study proposes a hybrid AI model consists of AIM, ANFIS, type-2 and artificial neural network (ANN) for predicting freeway traffic flow for the local freeway condition of Saudi Arabia.