عنوان

Applying Computational Intelligence Techniques to Forecast Traffic Flow Using Traffic Sensor Data & Weather Data

Number

TL49613

.Language of Text, Soundtrack etc

انگلیسی

Title Proper

Applying Computational Intelligence Techniques to Forecast Traffic Flow Using Traffic Sensor Data & Weather Data

General Material Designation

[Thesis]

First Statement of Responsibility

Danielle N. Clavon

Subsequent Statement of Responsibility

Islam, Muhammad F.

Name of Publisher, Distributor, etc.

The George Washington University

Date of Publication, Distribution, etc.

2018

Specific Material Designation and Extent of Item

92

Text of Note

Committee members: Eggstaff, Justin; Mazzuchi, Thomas A.; Rackley, Daphne; Sarkani, Shahram

Text of Note

Place of publication: United States, Ann Arbor; ISBN=978-0-355-47279-0

Dissertation or thesis details and type of degree

D.Engr.

Discipline of degree

Engineering Management

Body granting the degree

The George Washington University

Text preceding or following the note

2018

Text of Note

Traffic congestion is becoming a major problem in metropolitan areas across the globe. One useful way to attempt to mitigate traffic congestion is being able to forecast traffic flow. Traffic flow forecast must be accurate because of the critical part it plays in the development of intelligent transportation systems and SMART City initiatives for metropolitan areas. Many cities are in the process of deploying various technologies that range from traffic cameras to traffic signal cameras to improve the current state of traffic congestion as part of one of their SMART City initiatives. The era of Big Data for a number of cities is on the rise through all the new collection channels, which makes it critical to have statistical methods in place on how to interpret and analyze the new data. This praxis will focus on multivariate analysis. Sacramento, as well as other cities in California, will serve as a proxy for this praxis to illustrate the methodology. The praxis is designed to serve as a potential framework for other cities to adopt. The forecasts will be divided into two sections; AM Peak and PM Peak time. In order to aid in decreasing traffic congestions, an Artificial Neural Network was created to forecast traffic flow. The proposed methodology uses Levenberg Marquardt (LM) backpropagation for the Nonlinear Autoregressive Network with Exogenous Inputs (NARX) architecture. The dataset was collected from January 1, 2015, to August 31, 2016. The following variables were used for this study: flow, temperature, humidity, visibility, and speed. The results of the analysis proved that deploying NARX to forecast traffic flow is beneficial and provides an accurate forecast measured by Mean Absolute Percentage Error that ranges from 5% to 13% for the cities studied for this praxis. Therefore, the proposed methodology in the praxis can be applied to different cities in an effort to support their efforts of having the ability to forecast traffic flow to decrease congestion.

Computer Engineering; Transportation; Systems science

Subject Term

Applied sciences;Social sciences;Traffic flow;Traffic sensor data;Weather data

Alvi, Muzna Fatima

Islam, Muhammad F.

Subdivision

Engineering Management

The George Washington University

Call Number

1964722816; 10635715

Electronic name

p

[Thesis]

276903

a

Y