Short-Term Traffic Prediction with Vicinity Gaussian Process in the Presence of Missing Data

Abstract

This paper considers the problem of short-term traffic flow prediction in the context of missing data and other measurement errors. These can be caused by many factors due to the complexity of the large scale city road network, such as sensors not being operational and communication failures. The proposed method called vicinity Gaussian Processes provides a flexible framework for dealing with missing data and prediction in vehicular traffic network. First, a weighted directed graph of the network is built up. Next, a dissimilarity matrix is derived that accounts for the selection of training subsets. A suitable cost function to find the best subsets is also defined. Experimental results show that with appropriately selected subsets, the prediction root mean square error of the traffic flow obtained by the vicinity Gaussian Processes method reaches 18.9% average improvement with lower costs, which is with comparison to inappropriately chosen training subsets.

Bayesian Neural Networks Uncertainty Quantification with Cubature Rules

“…To deal with it, a number of numerical approximations of (1) are proposed, such as particle filtering [15], [16] and Gaussian regression [17]. Recently, the cubature rule [18] was applied to numerically approximate the integral, which has achieved at least third-order Taylor series accuracy [14].…”

SBAG: A Hybrid Deep Learning Model for Large Scale Traffic Speed Prediction

“…In input data , each sample is a 2-dimensional vector. The dimensions of input data are [ , ] = [10,323], based on model description. The time lag is set as 10. www.ijacsa.thesai.org…”

“…In general, regarding taking care of complex traffic forecasting problems [5], the computational approaches shattered the statistical methods like autoregressive integrated moving average (ARIMA) [6] in terms of capacity to catch nonlinear relationship and to deal with complex data. By the ascent of neural systems (NN) based methods, the full protentional of artificial intelligence was not subjugated in any case but many neural network-based models like Feed Forward Neural Network [7], Fuzzy Neural Network [8], Recurrent Neural Network [9], Gaussian Process [10] and hybrid Neural Network [11] [12] are adopted for traffic forecasting problems. Recently, some hybrid architectures are proposed for traffic speed forecasting.…”

A computationally efficient symmetric diagonally dominant matrix projection-based Gaussian process approach

“…However, they have cubic computational complexity O(n 3 ) for the inversion of the kernel matrix of size n × n and its determinant [3]. This cubic computational cost has effectively limited applications of GPs to data with thousands of samples [4,5]. This led to intensive studies of the scalability of GPs during the last decades [3], with particular interests in adapting GPs for various data processing and maintaining their capacity, ideally at the same level of full GPs.…”