Introduction:
Over the past decades, many researchers have shown great interest in the artificial-neural-networks (ANNs) and their applications in industry, business, as well as private and government sectors. Artificial neural networks as a sub-discipline of Artificial intelligence have recently emerged in the engineering world. Further, several researchers investigated the efficiency of using the ANN models in the transportation engineering field (Freedson et al., 2011).
Literature Review
Traffic congestion caused by accidents is considered one of the main traffic obstruction sources, especially in large urban cities in the United States. Over the last few years, various accident-warning algorithms have been proposed for freeway monitoring and control devices. Although traditional algorithms have met varying degrees of success in terms of performance parameters, the enhanced techniques are urgently needed, particularly with the introduction of intelligent transportation control systems. Moreover, intelligent transportation control systems will highly reliant on the ability to automatically detect traffic congestion. Artificial-neural-networks models (ANNs) were conducted to predict lane-blocking accidents on urban highways. The results indicate that ANN models have the potential to increase the efficiency of accident detection significantly (Ritchie and Cheu, 1993).
The Intelligent transportation network’s applications include short-term estimation of traffic patterns, such as motion and occupancy. Although several methodologies have been developed for short-term estimation, ANNs are one of the best options for modeling and forecasting traffic parameters. However, due to a lack of information about the ideal structure of a network particular dataset, researchers must depend on time-consuming and inefficient rules-of-thumb when creating the network. (Vlahogianni et al., 2005), established a multilayered structure strategy, based on an advanced genetic algorithm. This strategy can help with the proper assessment of traffic volume data with spatially and temporally features, as well as the implementation of the best ANN structure. Further, it assesses the established network’s efficiency, by applying it to multivariate traffic data.
(Kumar et al., 2013), investigated the traffic volume forecasting in the period term for non-urban highways, by creating an Artificial-Neural-Network model, and using old traffic data. The input parameters of the ANN model include the traffic volume, density, duration, and speed, along the day of the week. Further, the speed of each class of vehicles was separately adopted as input parameters, in contrast to previous research, that adopted the average traffic speed flow. The results clearly showed that the ANN model could predict the number of vehicles accurately, even when vehicle classes and their speeds were treated separately as input parameters.
The type of vessel incident can be predicted with a high degree of accuracy. Generating ANN models with input parameters such as weather conditions, a specific location, time, and traffic volume might reduce marine disasters. Thus, the official’s navigation and port authorities can be notified of the possibility of a particular accident. (Hashemi et al., 1995), investigated three models to forecast vessel incidents on the lower-Mississippi-River: the neural network model, regression analysis model, and finally the discriminant analysis model. The results showed that about 53% and 56% of injuries were correctly analyzed using discriminant and regression analysis. Moreover, these injury categorized into three main groups: crashing, grounding, in addition to ramming (Hashemi et al., 1995).
2021-4-28-1619618452