Hidden Markov Model Applied for Vehicles Density Prediction

Abstract

The Gempol-Pandaan toll is a strategic center for regional growth. Number of vehicles passing through this toll has increased in 2017. Because of the passing of numerous cars, particularly trucks carrying loads above their capacity, this condition may result in traffic congestion and road damage. As a result, it is essential to predict the number of vehicles coming from not only the Gempol toll gate but also the Kejapanan, Bangil, and Rembang toll gates that will exit through the Pandaan toll gate. The probability and quantity of each vehicle category—I, II, III, IV, and V—are used in this study to analyse vehicle density. The origin gate and vehicle category cannot be observed directly, but the car must pass the toll gate in order to tap the e-toll card, so the Hidden Markov Model (HMM) method was used in this study. The Expectation-Maximization (EM) algorithm and the Bayesian approach are the two estimation techniques used in this study for HMM parameters. The outcome demonstrates that Bayesian parameter estimation for HMM is superior to the EM algorithm. The model is more representative to explain the predicted vehicle density because the Bayesian estimated parameter values are nearer to the input parameters.