A hybrid algorithm for estimating origin-destination flows

With the development of intelligent transportation systems, the estimation of traffic flow in urban areas has attracted a great attention of researchers. The timely and accurate travel information of urban residents could assist users in planning their travel strategies and improve the operational efficiency of intelligent transportation systems. Currently, the origin-destination (OD) flows of urban residents are formulated as an OD matrix, which is used to denote the travel patterns of urban residents. In this paper, a simple and effective model, called NMF-AR, is proposed for predicting the OD matrices through combining the nonnegative matrix factorization (NMF) algorithm and the Autoregressive (AR) model. The basic characteristics of travel flows are first revealed based on the NMF algorithm. Then, the nonlinear time series coefficient matrix, extracted from the NMF algorithm, is estimated based on the AR model. Finally, we predict OD matrices based on the estimated coefficient matrix and the basis matrix of NMF. Extensive experiments have been implemented, in collected real data about taxi GPS information in Beijing, for comparing our proposed algorithm with some known methods, such as different kinds of $K$-nearest neighbor algorithms, neural network algorithms and classification algorithms. The results show that our proposed NMF-AR algorithm have a more effective capability in predicting OD matrices than other models.