Darmawan Satyananda, Sarina Sulaiman, Nur Haliza Abdul Wahab, Nurulhuda Firdaus Mohd Azmi
Many deep learning (DL) models have been proposed to predict future traffic conditions based on historical data, with their respective achievements, strength, and weakness. However, they only rely on traffic state and ignoring other influencing states and only learn recent traffic pattern. To improve accuracy of prediction, some improvements are proposed. One of them is utilization of external factors, because the traffic flow is not only influenced by traffic flow data only but also by various external factors. This article dis-cusses involvement of external factors in improving accuracy of traffic fore-casting. Three DL models are compared to evaluate the achievements in four different datasets. In general, the result is that involvement of external factor could improve traffic forecasting accuracy. In addition, not only the methods that take role in making accuracy, but also characteristic of the data has significance in the accuracy. © 2024 American Institute of Physics Inc.. All rights reserved.
Faculty of Computing, Universiti Teknologi Malaysia, Johor, Johor Bahru, 81310, Malaysia; UTM Big Data Centre, Soft Computing Research Group, Faculty of Computing, Universiti Teknologi Malaysia, Johor, Johor Bahru, 81310, Malaysia; Advanced Informatics School, Razak Faculty of Technology and Informatics, Universiti Teknologi Malaysia, Johor, Johor Bahru, 81310, Malaysia; Fakultas Matematika dan Ilmu Pengetahuan Alam, Universitas Negeri Malang, Jalan Semarang 5, East Java, Malang, 65145, Indonesia