Machine Learning Based on Neural Network Fitting for Traffic Behaviour at Road Intersection in Mixed Traffic Condition

Abstract

Fajaruddin Mustakim, Azlan Abdul Aziz, Lim Heng Siong, Yaser Bakhuraisa, Othman Che Puan, Mohammad Nazir Ahmad, Rabiah Abdul Kadir and Riza Sulaiman

Artificial Neural Networks (ANNs) play an essential role in artificial intelligence to explore and simulate the traffic behaviour on road network safety. In this study, eight cluster as input variables and one output were utilize to simulate the performance of the model. Input predictors involved traffic of conflict, vehicle category, second vehicles passing right turn motor vehicles (RMV), first vehicles passing (RMV), speed limit, gap pattern, day time, and infrastructure. Meanwhile output variables were right turn motor vehicles (RMV). Neural Network Fitting apply as the Machine Learning has been implemented to measure the mean square error and the regression value. The network was trained with eight hundred and forty-one datasets has been collected on mix traffic condition. Neural Network Fitting consist three approaches to trained the datasets namely Levenberg- Marquardt Algorithm (LMA), Bayesian-Regularization Algorithm (BRA) and Scaled Conjugate Gradient Algorithm (SCGA). Two layer feedforward network were use to analyse the regression. The assessment between those machine learning is carried out to justify the best performance outcomes. This study reveals that scaled conjugate gradient algorithm perform the best result in training, validating and test process for mean square error less than 0.05 meanwhile regression value determines more than 0.90.

PDF