Traffic Flow Forecasting Models for Intelligent Transportation System

ENGINEERING PROJECT PREPARATION
By Name
Course
Instructor
Institution
Location
Date

ABSTRACT
In the intelligent transportation system, the capacity to forecast traffic volume has been
considered very important. Traffic volume forecast will definitely contribute to the proactive
control of traffic and exact level of time estimation. It is important to note that the previous
attempts to have such projects succeed have been met with a number of challenges.
This particular study seeks to develop similar models for two sites in the capital city of Nairobi.
Four models were developed and later tested for one particular problem of forecasting. This was
defined as estimation of traffic flow for a period of 15 minutes. The four models that were
considered in the study included time series, historical average, on parametric progression and
finally neural network. In the end of the entire study, it was found that the nonparametric
regression model outperformed other models.

INTRODUCTION
Nairobi and the nation have been facing challenges of transportation of increasing magnitude and
also complexity. Due to the matters that are related to the financial constraints, construction of
facilities and enlargement of already existing ones have become very less attractive. There is an
urgent need to have efficient and very safe system of transportation. This improvement will be
effectively being accomplished through use of intelligent transportation system (ITS).ITS can be
defined as the application of technology of information in the sector of transportation. The
fundamental objective of ITS is to assist in the provision of environment that enables application
of improved decision making on transport system(Chan, Dillon and Chang 2013).
In some cases the traffic agencies can use this system to control signals of traffic and also to
ease congestion in the city that normally results into accidents. Similarly, travellers can use IT’S
in making informed decisions in regard to the travelling time, mode of travel and routes to be
used. In summary, the system will allow society to utilize system of surface transportation more
intelligently leading to safer and more efficient travel.
Although there are lots of benefits of this system, there are risks that are associated with the
same. One of such difficulties includes use of a wide variety of sensors and media of
communication that supports collection of data and subsequent application as already found in
the market. The software systems that are used in the analysis and processing of data have not
fully advanced. Such systems are very important since they provide real intelligent in the ITS.As
specialist of traffic control may not be expected to do a selection of optimal timing of signals that
are based on improper vehicle counts( Zhang and Haghani 2014).

Also, travellers cannot be expected to properly identify an optimal route to the destination as
indicated by such counts. It is very obvious that advanced tools of analysis which may include
artificial intelligence, optimization and simulation may be required. The traffic flow prediction
provides the ability to estimate future traffic volume that is measured in the units of vehicles per
hour.
PURPOSE AND AIM
The main aim of this study was to investigate the feasibility of forecasting freeway traffic and to
construct a structure that supports use of forecasting capability in the traffic management and
services of traveller information. All the data used to construct and support evaluation of
candidate forecasting model were collected at the department of transport. This was done so as to
ensure that the developed model becomes compatible with the future movement systems of
traffic and also with the existing traffic systems.
LITERAURE REVIEW
Previous attempts at Traffic Flow Forecasting
The process of traffic flow forecasting has had unsatisfying attempts in the research history.
Most of these trials have been using signals to develop in the control of systems. Urban traffic
control system is one of such cases. In this particular system there is limitation on the number of
the freeway traffic flow. Prediction of traffic is done using different approaches that are largely
dependent on the well-defined patterns. The classification of the previous traffic flow can be
done using different cartegories.These categories are time series models and historical data based
algorithms(Xia et al 2014).