With the help of huge traffic data available from sensors, predictions of real-time or near-future traffic conditions are possible nowadays through various traffic flow prediction models (TFPM). Predictions of traffic flow are considered as the foundation of many innovative intelligent transportation systems (ITS) approaches. With the application of TFPM, ITS plays a major role in urban traffic management to reduce congestion and delays. One such approach of ITS is advanced traffic management systems (ATMS). Hence with ATMS strategies, centralized systems are developed to control and maintain stable traffic in urban areas. Traffic flow predictions are also a part of advanced traveller information systems (ATIS) that helps the traveller to do a better route choice and hence reduce travel time and delays. Knowing the importance of predicting urban traffic conditions, this paper will work on traffic flow prediction models that forecast traffic flow at a busy urban intersection in Duisburg, Germany. Traffic flow data were collected from inductive loop detectors and were used to train and test the models. Many research works have been done based on conventional statistical models and modern deep learning models for data-based time series forecasting. After the recent development of the Fb-Prophet and Neural Prophet models, there is a need to check their performance by comparing them with traditional models. One of the precise statistical models is the Seasonal Autoregressive Integrated Moving Average model (SARIMA). This paper aims to forecast traffic flow with Fb-Prophet and Neural Prophet models and tries to compare their performance with the SARIMA model in terms of data pre-processing, accuracy (MAPE) and computational requirements.
Comparison of SARIMA, Fb-Prophet and Neural Prophet Models for Traffic Flow Predictions at a Busy Urban Intersection
Lecture Notes in Civil Engineering
International Conference on Sustainable Infrastructure: Innovation, Opportunities and Challenges ; 2023 ; Mangalore, India April 23, 2023 - April 24, 2023
2024-09-06
17 pages
Article/Chapter (Book)
Electronic Resource
English
Elsevier | 2023
|Prophet-DCRNN traffic flow prediction method fusing multi-modal information
European Patent Office | 2021
|Traffic volume prediction method of Prophet-DeepAR model
European Patent Office | 2023
|