A method for repairing missing traffic data based on Bayesian enhanced tensor

Abstract

The invention discloses a method for repairing missing traffic data based on Bayesian enhanced tensor decomposition. The steps are as follows: organize road network vehicle speed data into a third-order data tensor and introduce a dominant factor structure for modeling; input data tensor , indicates a tensor; update the posterior distribution of the global parameter μ; update the posterior distribution of the hyperparameter; update the posterior distribution of the bias parameter φ and the posterior distribution of the factor matrix parameter U until i = m; update the bias parameter The posterior distribution of θ and the posterior distribution of factor matrix parameter V, until j=n; Update the posterior distribution of bias parameter η and the posterior distribution of factor matrix parameter X, until t=f; Repeat steps S4~S9 until The difference between the precision parameter τ and the parameter τ of the previous generation Δτ <ε时，则模型收敛，进入下一步；利用更新后的{μ,φ,θ,η,u,v,x}参数值，代入yCalculate the estimated tensor in the ijt expression< / ε时，则模型收敛，进入下一步；利用更新后的{μ,φ,θ,η,u,v,x}参数值，代入y

Description

technical field [0001] The invention relates to the technical field of intelligent traffic systems, and more specifically, to a method for repairing missing traffic data based on Bayesian enhanced tensors. Background technique [0002] Missing data has become a common and unavoidable problem in the field of intelligent transportation systems, and the reasons for this problem are various. First of all, due to the natural sparsity of some traffic data, it cannot be effectively collected completely. In addition, the limited spatial distribution of sensors limits the completeness of data from the perspective of traffic management costs. Furthermore, uncertainty factors such as communication failure and transmission failure of the data acquisition equipment itself are another common factor. Therefore, it is necessary to accurately repair missing data and enhance data quality to support the application of intelligent transportation systems. [0003] Taking road speed as an exam...

AI Technical Summary

Problems solved by technology

Existing traffic data filling techniques based on tensor decomposition cannot simultaneously mine explicit and implicit traffic features, resulting in inaccurate data estimation and poor model interpretation

Images