Construction of traffic state vector using mutual information for short-term traffic flow prediction

Fuzzy hybrid framework with dynamic weights for short‐term traffic flow prediction by mining spatio‐temporal correlations

“…max-relevance), and (ii) each selected variable should have little redundancy with any other variable in the selected variable subset (i.e. minredundancy) [48][49][50]. Traditional traffic feature construction usually constructs the state vector by directly taking the traffic flow values as the feature values, several intervals before the interval to be predicted.…”

“…In this study, the historical state vector corresponding to the previous periods of predicted time (M h (t − 1)) was added to form a new vector matrix [6]. For a specific location in the space domain, the observed values of traffic flow may be affected by the upstream (M up h (t)) and downstream (M down h (t)) flow of the observation point [49]. We further expanded the state vector matrix by adding the corresponding state vector matrix of the upstream and downstream observation points.…”

An Improved Selective Ensemble Learning Method for Highway Traffic Flow State Identification

“…Ryu et al adapted a K-Nearest Neighbor model for the application of the proposed state vector and proposed a method for constructing traffic state vectors by using mutual information. The experimental results for realworld traffic data show that the proposed method of constructing a traffic state vector provides reasonable prediction accuracy in short-term traffic prediction [16]. Su et al presented a traffic state forecasting method using adaptive neighborhood selection that is based on an expansion strategy to search manifold neighbors and obtain higher precision with manifold neighbors.…”

Prediction of Traffic Flow via Connected Vehicles

“…To this end, predictive techniques are needed by infrastructure operators to allow advanced modelling. The real-time prediction of traffic flow on a road segment allows transportation authorities to take actions to control traffic load and prevent congestion [1], [2]. Particularly, Short-Term traffic Prediction (STP) enables traffic managers to take informed decisions about how to best reroute traffic, change lane priorities and modify traffic light timing.…”