International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 02 | Feb 2025
p-ISSN: 2395-0072
www.irjet.net
Neural Network-Driven Road Traffic Forecasting for Optimized Smart Network Infrastructure Manju J1, Sumalatha M S2 1Professor, Mahaguru Institute of Technology, Kayamkulam
2Associate Professor, Department of computer science and engineering, Mahaguru Institute of Technology,
Kayamkulam ---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Road traffic forecasting is the process of
road parts. Because the TSE topic is useful to public authorities, road workers, and the general people, there is a lot of interest in investigating it in the ITS field [1].
predicting future traffic conditions and patterns using data analysis. It is crucial for the better traffic management, safety enhancement, congestion reduction, and urban mobility. Current road transport forecasting approaches are limited by inaccuracies in real-time data, poor handling of complex traffic dynamics and unpredictable human behavior. In order to forecast traffic flows using radar and meteorological sensor data, this study evaluates the efficacy of three advanced deep learning approaches: convolutional neural networks (CNN), long short-term memory (LSTM) and autoregressive LSTM (AR-LSTM). The system demonstrated that they accurately predict traffic patterns by including weather as an essential element. The study shows how weather has a big influence on traffic flow, which makes it crucial for predictive modeling. In terms of computational efficiency and prediction accuracy, the CNN model performed better than both LSTM and AR LSTM. CNN demonstrated its suitability for real-time applications by achieving the lowest mean absolute error and require a small amount of execution time to create predictions. The study demonstrates the potential of these models to assist traffic management, which shows that it is feasible to estimate traffic flows with a high degree of accuracy across one-hour intervals. The importance of implementing advanced deep learning techniques for more intelligent and responsive traffic systems is highlighted by these findings.
Road traffic situation problems are a typical occurrence in all major cities globally. For more than 50 years, governmental and private organizations have worked to reduce the social, economic, and environmental issues caused by traffic congestion. Three methods used to lessen the effects of traffic congestion: controlling traffic flows, promoting transportation alternatives, and building more infrastructure. The second is primarily a subject of public policies, while the first is constrained by topographical, financial, and social factors. The latter has been improving steadily over the past few years due to the data growth provided by sensors in vehicles and roads, as well as the technology required for using that data. In order to create advanced traveller information systems (ATIS) and advanced traffic management systems (ATMS), it is helpful to be able to monitor, analyze, and understand traffic factors like flow, occupancy, or travel times. Early investigations of Kalman filtering techniques and time-series approaches using various approaches comprised the majority of the initial efforts to predict traffic flows. Figure 1 depicts the road traffic forecasting model.
Key Words Road traffic forecasting, convolutional neural networks, mean absolute error, deep learning.
1.INTRODUCTION Road traffic forecasting can optimize route planning, enhance decision-making, and improve overall transportation efficiency. The rise in traffic congestion is one of the main problems that drivers, highway operators, and municipal managers are currently dealing with. It worsens the traveller experience by adding complexity to everyday travel and having harmful effects on the environment, users' time, and financial expenses. Road and city operators may also find it useful to use traffic flow forecasts when putting traffic planning and management techniques into practice. Traffic state estimation (TSE) is the technique of using noisy and partially observed traffic data to infer traffic state parameter such as speed, flow, and other similar factors, on
© 2025, IRJET
|
Impact Factor value: 8.315
Fig – 1: Road traffic forecasting model Short-term prediction horizons have been used in many methods to forecast traffic features. The research community's interest in this area has grown considerably, as have the availability of data, analysis techniques, and processing power. Predicting traffic details is one of the
|
ISO 9001:2008 Certified Journal
|
Page 357