HTFM: Hybrid Traffic-Flow Forecasting Model for Intelligent Vehicular Ad hoc Networks

Short-Term Traffic Prediction Using Deep Learning Long Short-Term Memory: Taxonomy, Applications, Challenges, and Future Trends

“…Therefore, an approach is adopted in which the ConvNNs capture the inter and intra-day patterns of traffic, which are fed to the LSTM to acquire traffic features in time and then performs the traffic forecast. In [58], a hybrid traffic-flow forecasting model (HTFM) is proposed in which the spatio-temporal characteristics of the vehicles data flow on a large scale are addressed. The time characteristics are learned by the LSTM mechanism while the space characteristics are obtained by the ConVNNs.…”

“…Complexity of processing in switching between post and pre data processing phases [51] Considers RNNs for traffic flow modeling and uses the LSTM to deal with the vanishing gradient issue in data prediction supported by a learning mechanism Improved prediction accuracy and ease of training the LSTM with the attention mechanism RNNs tend to lose information weight for high data streams [55] Data features are identified in time and space and periodic data sequences are assigned different weights followed by processing previous and the newly generated data values for forecasting Ease of traffic flow forecasting based on periodicity of data flow Processing delay due to computation of previous and newly computed data values (bidirectional) [56] Predicts traffic for multi-lanes by considering routing among the vehicles in the lanes, the bidirectional traffic and recursive data processing Promising forecasting for multi-lane highways Data congestion/interference due to multi-lanes traffic and routing [57] Uses inter and intra-day traffic patterns with extracted features and their correlation with the weather conditions 90 % prediction accuracy Acquisition of extra intra-days information is required [58] Obtains the features of a large scale traffic flow in space and time and analyzes them to perform data traffic prediction Accurately predicts data flow patterns due to their time and space characteristics on a large scale Processing large scale data increases resources utilization and delays the real time implementation necessary for congested highways [59] First obtains the space and time features of the traffic and then uses the maximum information coefficient to correlate the extracted features and predict the parameters Conveniently detects and predicts traffic fluctuations Lacks in time-efficiency for dense traffic conditions with rapid fluctuations [60] Correlates the traffic transfer patterns among roads and their spatio-temporal characteristics for traffic flow prediction…”