Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

Accurate and timely traffic flow information is important for the successful deployment of intelligent transportation systems. Over the last few years, traffic data have been exploding, and we have truly entered the era of big data for transportation. Existing traffic flow prediction methods mainly use shallow traffic prediction models and are still unsatisfying for many real-world applications. This situation inspires us to rethink the traffic flow prediction problem based on deep architecture models with big traffic data. In this paper, a novel deep-learning-based traffic flow prediction method is proposed, which considers the spatial and temporal correlations inherently. A stacked autoencoder model is used to learn generic traffic flow features, and it is trained in a greedy layerwise fashion. To the best of our knowledge, this is the first time that a deep architecture model is applied using autoencoders as building blocks to represent traffic flow features for prediction. Moreover, experiments demonstrate that the proposed method for traffic flow prediction has superior performance.

Traffic Flow Prediction With Big Data: A Deep Learning Approach

The development of the Internet in recent years has made it possible and useful to access many different information systems anywhere in the world to obtain information. While there is much research on the integration of heterogeneous information systems, most commercial systems stop short of the actual integration of available data. Data fusion is the process of fusing multiple records representing the same real-world object into a single, consistent, and clean representation.This article places data fusion into the greater context of data integration, precisely defines the goals of data fusion, namely, complete, concise, and consistent data, and highlights the challenges of data fusion, namely, uncertain and conflicting data values. We give an overview and classification of different ways of fusing data and present several techniques based on standard and advanced operators of the relational algebra and SQL. Finally, the article features a comprehensive survey of data integration systems from academia and industry, showing if and how data fusion is performed in each.

Data fusion

Highway Capacity Manual改訂の動向--テイラ-教授の講演より

Gaining a better understanding of the factors that affect the likelihood of a vehicle crash has been an area of research focus for many decades. However, in the absence of detailed driving data that would help improve the identification of cause and effect relationships with individual vehicle crashes, most researchers have addressed this problem by framing it in terms of understanding the factors that affect the frequency of crashes - the number of crashes occurring in some geographical space (usually a roadway segment or intersection) over some specified time period. This paper provides a detailed review of the key issues associated with crash-frequency data as well as the strengths and weaknesses of the various methodological approaches that researchers have used to address these problems. While the steady march of methodological innovation (including recent applications of random parameter and finite mixture models) has substantially improved our understanding of the factors that affect crash-frequencies, it is the prospect of combining evolving methodologies with far more detailed vehicle crash data that holds the greatest promise for the future.

http://ceprofs.tamu.edu/dlord/Papers/Lord-Mannering_Review.pdf

The statistical analysis of crash-frequency data: A review and assessment of methodological alternatives

After reviewing the problem of short-term traffic forecasting, a nonparametric regression method, the \Ik\N-nearest neighbor (\Ik\N-NN) approach, is suggested as a candidate forecaster that might sidestep some of the problems inherent in parametric forecasting approaches. An empirical study using actual freeway data is devised to test the \Ik\N-NN approach, and compare it to simple univariate linear time-series forecasts. The \Ik\N-NN method performed comparably to, but not better than, the linear time-series approach. However, further research is needed to delineate those situations where the \Ik\N-NN approach may, or may not be, preferable. Particular attention should be focused on whether or not regression methods, which forecast mean values, are appropriate for forecasting the extreme values characteristic of transitions from the uncongested traffic regime to the congested regime. In addition, larger data bases may improve the accuracy of the \Ik\N-NN method.

Nonparametric Regression and Short‐Term Freeway Traffic Forecasting

The application of recursive prediction error techniques to the problem of estimating origin-destination patterns from input and output volume counts is described. Each algorithm deals with the special case where route choice between origin and destination can be ignored. A gradient algorithm developed by Cremer and Keller (1983) turns out to be a special case of a family of methods described by Ljung and Soderstrom (1983). After describing how the methods developed in Ljung and Soderstrom (1983) could be modified so that the resulting estimates satisfy natural constraints, a number of algorithm possibilities are tested. Generally, those algorithms employing Gauss-Newton search directions appear superior to gradient-based methods, while the constraining procedures improve accuracy.

Recursive estimation of origin-destination matrices from input/output counts

The continuous Network Design Problem (NDP) deals with determining optimal expansions for the capacities of a street network, subject to the constraint that the street traffic volumes must be the outcome of a user-optimal equilibrium assignment. Although the use of deterministic equilibrium methods tends to produce computationally intractable problems, in this paper it is shown that a stochastic user equilibrium based on the logit model leads to a differentiable and large-scale, but tractable, version of the NDP. A procedure for computing the derivatives of the stochastic user equilibrium (SUE) assignment without having to first compute the route choice probabilities is given, and this procedure is coupled with two standard algorithms for solving nonlinear programs, the generalized reduced gradient method and sequential quadratic programming. These algorithms are tested on several example networks, and the results of these tests suggest that the SUE-constrained version of the NDP offers both a promising heuristic for solving DUE-constrained problems as well as a viable procedure in its own right.

Exact local solution of the continuous network design problem via stochastic user equilibrium assignment

An ordered, discrete outcome model that relates the severity of injury suffered by a struck pedestrian to the speed of the striking vehicle is derived and then fit to previously published data. Particular care is taken to account for covariate measurement error and for the fact that the data were collected using an outcome-based, or retrospective, sampling plan. The results show similar patterns for children (ages 0-14) and adults (ages 15-59), but for elderly pedestrians (ages 60+) the injuries produced in crashes that involved lower impact speeds tended to be more severe than for the other two groups. Use of the model is illustrated by applying it to two reconstructed vehicle-pedestrian crashes to determine the likely severity of the injury had the driver adhered to a speed limit. Some implications for speed limits on residential streets are also pointed out.

Relating Severity of Pedestrian Injury to Impact Speed in Vehicle-Pedestrian Crashes: Simple Threshold Model

The use of prediction error and maximum likelihood techniques to estimate intersection turning and through movement probabilities from entering and exiting counts is considered here. A maximum likelihood estimator for situations when full information on turning movement counts is available is derived and used as a component for a maximum likelihood algorithm which only requires entering and exiting counts. Several algorithms based on minimizing the error between observed and predicted exiting counts are also developed. Some actual traffic data are collected and used to develop realistic simulations for evaluating the various estimators. Generally, the maximum likelihood algorithm produced biased but more efficient estimates, while prediction error minimization approaches produced unbiased but less efficient estimates. Constraining the recursive version of the ordinary least-squares estimator to satisfy natural constraints did not affect its long run convergence properties.

Gary A. Davis

Papers

Nonparametric Regression and Short‐Term Freeway Traffic Forecasting

Recursive estimation of origin-destination matrices from input/output counts

Exact local solution of the continuous network design problem via stochastic user equilibrium assignment

Relating Severity of Pedestrian Injury to Impact Speed in Vehicle-Pedestrian Crashes: Simple Threshold Model

Application of Prediction-Error Minimization and Maximum Likelihood to Estimate Intersection O-D Matrices from Traffic Counts