A Short-Term Traffic Flow Combination Prediction Model with Adaptive Weights
Chuanxia Sun, Xiaoliang Sun, Peixuan Yin, Shenyuan Zhang · CICTP 2020 · 2020
A time-varying weighted combination model is proposed to predict the short-term traffic flow in the paper. The changed weights will much minimize the influence of random factors and generalize the application conditions. Firstly, the training data are classified into several categories according to the traffic situations, and LS-SVM is used to train the single local prediction model of each category; Secondly, the occurring possibility of different traffic situations is made as the weight of the corresponding local prediction model; and then the model is constructed with the linear weighted combination strategy. Finally, the experiment with the actual speed data shows that the method is effective.