Nonlinear Traffic Prediction as a Matrix Completion Problem with Ensemble Learning

Abstract

We focus on short-term traffic forecasting for traffic operations management. Specifically, we focus on forecasting traffic network sensor states in high-resolution (second-by-second). Most work on traffic forecasting has focused on predicting aggregated traffic variables, typically over intervals that are no shorter than 5 minutes. We develop a (big) data-driven methodology for forecasting sensor states in high-resolution. Our contributions can be summarized as offering three major insights: first, we show how the forecasting problem can be modeled as a matrix completion problem. Second, we employ a block-coordinate descent algorithm and demonstrate that the algorithm converges in sub-linear time to a block coordinate-wise optimizer. This allows us to capitalize on the "bigness" of high-resolution data in a computationally feasible way. Third, we develop an ensemble learning (or adaptive boosting) approach to reduce the training error to within any arbitrary error threshold. The latter utilizes past days, so that the boosting can be interpreted as capturing periodic patterns in the data. The performance of the proposed method is analyzed theoretically and tested empirically using a real-world high-resolution traffic dataset from Abu Dhabi, UAE. Our experimental results show that the proposed method outperforms other state-of-the-art algorithms.

Learning Traffic Speed Dynamics from Visualizations

“…However, these methods are restricted by the representational power of the selected traffic flow model, and require additional inputs such as initial and boundary conditions and traffic flow model parameters. Furthermore, their estimations are often coarse grained, i.e., most studies assume an estimation interval of 100 − 500 m × 5 − 15 minutes (with an exception of a few recent studies [13]) and are not suitable for arbitrary space-time resolutions. The latter is a drawback of the numerical scheme used in solving the traffic flow model; for e.g., in the Eulerian coordinate system, the mesh sizes in the space and time dimensions are restricted by the Courant-Fredich-Lewy (CFL) condition [20].…”

Incorporating Kinematic Wave Theory into a Deep Learning Method for High-Resolution Traffic Speed Estimation

“…The most common type of machine learning tool used in data-driven approaches is deep learning (DL), specifically deep neural networks (DNNs) [5,12,20,24,34,40,56]. Other techniques include support vector machines [57], case based reasoning [21], random forests [13], decision trees [35], gradient boosting [65], kernel regression [59], principal component analysis [26], and matrix factorization methods [29,30]. The estimation results from data-driven methods are often reported to be more accurate than model-based approaches.…”