Data Mining Practical Machine Learning Tools and Techniques

1. Introduction and Overview. 2. Detecting Influential Observations and Outliers. 3. Detecting and Assessing Collinearity. 4. Applications and Remedies. 5. Research Issues and Directions for Extensions. Bibliography. Author Index. Subject Index.

Regression Diagnostics: Identifying Influential Data and Sources of Collinearity

to be done in this area. Face recognition is a problem that scarcely clamoured for attention before the computer age but, having surfaced, has involved a wide range of techniques and has attracted the attention of some fine minds (David Mumford was a Fields Medallist in 1974). This singular application of mathematical modelling to a messy applied problem of obvious utility and importance but with no unique solution is a pretty one to share with students: perhaps, returning to the source of our opening quotation, we may invert Duncan's earlier observation, 'There is an art to find the mind's construction in the face!'.

Linear and nonlinear waves, by G. B. Whitham. Pp.636. £50. 1999. ISBN 0 471 35942 4 (Wiley).

This article shows how the evolution of multi-commodity traffic flows over complex networks can be predicted over time, based on a simple macroscopic computer representation of traffic flow that is consistent with the kinematic wave theory under all traffic conditions. The method does not use ad hoc procedures to treat special situations. After a brief review of the basic model for one link, the article describes how three-legged junctions can be modeled. It then introduces a numerical procedure for networks, assuming that a time-varying origin-destination (O-D) table is given and that the proportion of turns at every junction is known. These assumptions are reasonable for numerical analysis of disaster evacuation plans. The results are then extended to the case where, instead of the turning proportions, the best routes to each destination from every junction are known at all times. For technical reasons explained in the text, the procedure is more complicated in this case, requiring more computer memory and more time for execution. The effort is estimated to be about an order of magnitude greater than for the static traffic assignment problem on a network of the same size. The procedure is ideally suited for parallel computing. It is hoped that the results in the article will lead to more realistic models of freeway flow, disaster evacuations and dynamic traffic assignment for the evening commute.

The cell transmission model, part ii: network traffic

Although the “first order” continuum theory of highway traffic proposed by Lighthill and Whitham (1955) and Richards (1956)—the LWR model—can predict some things rather well, it is also known to have some deficiencies. In an attempt to correct some of these, “higher order” theories have been proposed starting in the early 70s. Unfortunately, the usefulness of these improvements can be questioned. This note describes the logical flaws in the arguments that have been advanced to derive higher order continuum models, and shows that the proposed high order modifications lead to a fundamentally flawed model structure. The modifications can actually make things worse.

As an illustration of this, it is shown that any continuum model of traffic flow that smooths out all discontinuities in density will predict negative flows and negative speeds (i.e., “wrong way travel”) under certain conditions. Such unreasonable predictions are made by all existing models formulated as a quasilinear system of partial differential equations in speed, density, and (sometimes) other variables but not by the LWR model.

The note discusses the available empirical evidence and ends with a (hopefully positive) commentary on what can be accomplished with first-order models.

Requiem for second-order fluid approximations of traffic flow

In this work a functional form for the speed-density relationship is presented This functional form is made up of a nondimensional spacing, the equivalent spacing and of a function, the generating function, whose argument is the equivalent spacing This functional form is derived by means of two different arguments The first argument is based on the set of properties that the volume-speed-density relationships should satisfy The second one arises when applying a dimensional analysis to a generic car-following model Finally, several examples of generating functions are shown

On the functional form of the speed-density relationship—I: General theory

Acquiring high-quality origin-destination (OD) information for traffic in a geographic area is both time consuming and expensive while using conventional methods such as household surveys or roadside monitoring. These methods generally present only a snapshot of traffic situation at a certain point in time, and they are updated in time intervals of up to several years. A technique was developed that makes use of the global system for mobile communications (GSM) mobile phone network. Instead of monitoring the flow of vehicles in a transportation network, the flow of mobile phones in a cell-phone network is measured and correlated to traffic flow. This methodology is based on the fact that a mobile phone moving on a specific route always tends to change the base station nearly at the same position. For a first pilot study, a GSM network simulator has been designed, where network data can be simulated, which is then extracted from the phone network, correlated, processed mathematically and converted into an OD matrix. Primary results show that the method has great potential, and the results inferred are much more cost-effective than those generated with traditional techniques. This is due to the fact that no change has to be made in the GSM network, because the information that is needed can readily be extracted from the base station database, that is the entire infrastructure needed is already in place

http://www.esi2.us.es/GT/docs/iet_art1.pdf

Deriving origin destination data from a mobile phone network

In this part we applied the theory developed in Part I. Some speed-density curves having the functional form proposed in Part I are fitted to traffic data. The goodness of fit is excellent, except in the case of the left lane detectors. A procedure for isolating stationary traffic periods is also explained. This procedure satisfactorily eliminates the dispersion of the individual measurements.

On the functional form of the speed-density relationship--ii: empirical investigation

One of the main characteristics of modern society is the never-ending increase in mobility. This leads to a series of problems such as congestion and increased pollution. To resolve these problems, it is imperative to have a good road network management and planning. To efficiently identify the characteristics of traffic in the road network, it would be necessary to perform a permanent monitorisation of all roadway links. This would involve an excessive cost of installation and maintenance of road infrastructure. Hence, new alternatives are required which can characterise traffic in a real time with good accuracy at an acceptable price. Mobile telephone systems are considered as a promising technology for the traffic data collection system. Its extensive use in converting its subscribers in a broad sample to draw information from phones becomes anonymous probes to monitor traffic. It is reviewed how to obtain parameters related to traffic from cellular-network-based data, describing methods used in existing simulation works as well as field tests in the academic and industrial field.

Review of traffic data estimations extracted from cellular networks

Traffic volume is a parameter used to quantify demand in transportation studies, and it is commonly collected by using on-road (fixed) sensors such as inductive loops, cameras, etc. The installation of fixed sensors to cover all roads is neither practical nor economically feasible; therefore, they are only installed on a subset of links. Cellular phone tracking has been an emerging topic developed and investigated during the last few years to extract traffic information. Cellular systems provide alternative methods to detect phones in motion without the cost and coverage limitations associated with those infrastructure-based solutions. Utilizing existing cellular systems to capture traffic volume has a major advantage compared with other solutions, since it avoids new and expensive hardware installations of sensors, with a large number of cellular phones acting as probes. This paper proposes a set of models for inferring the number of vehicles moving from one cell to another by means of anonymous call data of phones. The models contain, in their functional form, terms related to the users' calling behavior and other characteristics of the phenomenon such as hourly intensity in calls and vehicles. A set of intercell boundaries with different traffic background and characteristics were selected for the field test. The experiment results show that reasonable estimates are achieved by comparison with volume measurements collected by detectors located in the same study area. The motion of phones while being involved in calls can be used as an easily accessible, fast, and low-cost alternative to deriving volume data on intercell boundaries.

Francisco G. Benitez

Papers

On the functional form of the speed-density relationship—I: General theory

Deriving origin destination data from a mobile phone network

On the functional form of the speed-density relationship--ii: empirical investigation

Review of traffic data estimations extracted from cellular networks

Traffic Flow Estimation Models Using Cellular Phone Data