Fuzzy Logic Model For The Prediction of Traffic Volume in Week Days

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

International Journal of Computer Applications (0975 – 8887)

Volume 107 – No 17, December 2014

Fuzzy Logic Model for the Prediction of Traffic Volume in


Week Days
Bharti Sharma Vinod Kumar Katiyar Arvind Kumar Gupta
Department of IT, Department of Mathematics, Department of Mathematics
College of Engineering Indian Institute of Technology , Indian Institute of Technology,
Roorkee,Roorkee-247667, Roorkee, Roorkee-247667,
Ropar,
Uttarakhand, India Uttarakhand, India
Punjab -140001, India

ABSTRACT Finally, Section VI summarizes the paper and presents some


This paper presents a model for traffic volume prediction directions for future work
which can be effectively used for transportation planning,
management and security assessment at any time. Fuzzy logic
2. RELATED WORK
is applied in order to realize effective and efficient traffic Detection of traffic flow in transportation is a critical area of
prediction. In this paper, ‘day’ of a week and ‘time’ of a day research among researchers to assist the decision-making
are taken as inputs for proposed model and the output will be processes [5].Research is done by various researchers in the
the predicted the traffic volume. The ‘time’ is divided into field of transportation to predict traffic flow. Traffic Theory
nine triangular membership functions. The second input ‘day’ categorized all the forecasting models developed by
is divided into five triangular membership functions and the researchers into two categories one is mathematical model and
output forecasted traffic volume has been divided into eight second is knowledge-based intelligent model. The
triangular membership functions. The predicted traffic volume mathematical models include Historical average the Kalman
when compared with actual traffic volume has MAPE within Filter (KF) [6][7], Exponential Smoothing [8] methods and
acceptable level of error. Prediction results show that the knowledge-based intelligent models include Artificial Neural
proposed fuzzy logic system produces more accurate and Networks (ANN)[9][10][11][12] , Neuro-Fuzzy Systems [13]
stable traffic volume predictions. and Fuzzy[14] .The accuracy of the mathematical forecasting
method cannot satisfactorily meet the demand of real-time
Keywords traffic control systems [15] because the traffic system is a
complex and variable system that involves a great deal of
Defuzzification; Fuzzy logic; MAPE; Membership functions ;
people, the traffic flow state has high randomness and
Traffic load forecasting
uncertainty. The mathematical models of traffic flow forecast
1. INTRODUCTION methods such as Kalman filter, Moving Average (MA),
Autoregressive Integrated Moving Average (ARIMA) and etc
Nowadays intelligent transportation systems are being
had been unable to satisfy the demand of the forecast accuracy
developed however lot of difficulties such as traffic flow,
that was increasing in practice [16] .On the other hand,
parking, road accidents are not solved yet [1]. Traffic flow
knowledge-based intelligent model based on artificial neural
prediction accuracy is one of the critical tasks to improve
networks have been applied many problems because of their
traffic conditions in the urban streets. The method to improve
non-linear modeling capability .During the last decade
traffic flow prediction accuracy is still an important problem
artificial neural network have been applied continuously to
of intelligent transportation systems. There are two types of
predict the traffic data[17]. Different neural network methods
Traffic flow prediction. Short- term forecasting and Long-
like MLP [18], RBF [19], ERNN [20], Neuro-Genetic [21]
term forecasting respectively. Short- term traffic flow
have been applied to predict the traffic flow.The performance
forecasting defines the traffic intensity for the next time
of the artificial neural network mainly depends on the network
interval, usually from 5 to 30 minutes. Long- term forecasting
training and is extremely by the size and quality of training
of the traffic intensity can be predicted hours, days or even
data and different design parameters like the number of
years ahead.
hidden layers, number of epoch and the number of elements
Short-term traffic flow prediction has been related to in each hidden layers. In this paper traffic volume is predicted
intelligent transportation systems. Short term traffic flow for week days using fuzzy logic to get accurate and exact
prediction supports (1) the creation of traffic control methods traffic volume.
in advanced traffic management systems (ATMS)[2](2) real-
time route guidance in advanced traveler information systems 3. FUZZY LOGIC APPROACH FOR
(ATISs)[3] and(3) evaluation of different dynamic traffic TRAFFIC FLOW PREDICTION
control and guidance methods as well. In an early report on
the architecture of intelligent transportation systems[4] it was 3.1 Fuzzy logic
clearly mentioned that the potential to make predictions of Fuzzy logic has been mostly used to solve many problems in
traffic flows and link travel times for several minutes into the system identification, signal processing and control [22]–[23].
future, using real-time traffic data, is a critical requirement for In fuzzy systems research, Fuzzy modeling has been
providing dynamic traffic control and guidance. considered as one of the critical problems [24] [25]. In the
previous years, research concentrated on the growth of fuzzy
This paper is organized as follows. Section II describes related systems from both theoreitical and applications oriented
work. Section III provides methodology for proposed model prospective [26]–[28]. A set of fuzzy If-Then rules is used to
and Section IV describes the proposed model to predict traffic represent a fuzzy model structure [29].A fuzzy rule has two
volume using fuzzy logic. Results are presented in Section V. elements , antecedent and the consequence. The antecedent

1
International Journal of Computer Applications (0975 – 8887)
Volume 107 – No 17, December 2014

variables contain information about the process operating Delhi-hardwar highway (NH-58) road is used to ensure the
conditions. The consequent of the rule is shows a linear fuzzy sets of each variable and its corresponding membership
regression model which is valid for the given operating functions. Day of week is fuzzified into five fuzzy sets: Mon,
condition [30]–[34]. Consider a set of inputs x1, x2, . . . , xn Tue, Wed, Thur, Fri. The time is fuzzified into midnit, dawn ,
variables and y single output variable for dynamic system . emor , mor, fnoon , anoon , even , dusk , night. To get better
The relationship between input and output variables in prediction precision, the output variable (traffic volume) is
dynamic system can be represented as: fuzzified into eight sets: Very low, low, medium, moderate,
high, Vhigh, v2high, v3high.Triangular function is used to
y = f(x1, x2, . . . , xn) (1) denote membership function for the input and output
A fuzzy model structure describes the function f between the variables. Figure 2a, 2b and Table 1(a) and Table 1(b) are
inputs xi , i = 1, . . . , n and the output y. The inputs and used to represent memberships function of the input variables
outputs are measured at the sample time t; t = 1,2, . . . , N. and figure 3 ,Table 3 for the membership function of output
variable .
A fuzzy model of a dynamic system for a Many Input Single
Output (MISO) system has a set of rules in the following Table 1: Parameters of membership function of input
format: Variables
Rulek : If x1 is Dk1 and . . . and xn is Dkn then
(a) Day of the week
yk = dk1x1 + dk2x2 + · · · + dknxn + dk0 Mon [1,1,1,1]
(2)
Tue [2,2,2,2]
where k is the rule number. For example, if we want to model
a three input n = 3 single output system using two rules i = 1, Wed [3,3,3,3]
2. The system can be represented as follows:
Thur [4,4,4,4]
Rule1: If x1 is D11 and x2 is D12 and x3 is D13 then
Fri [5,5,5,5]
y1 = d11x1 + d12x2 + d13x3 + d10
(b) Time of the Day
Rule2: If x1 is D21 and x2 is D22 and x3 is D23 then
Midnit (0,0,3)
y2 = d11x1 + d12x2 +da13x3 + d20
where y1 is the output corresponding to the region of values, Dawn (0,3,5)
in the input domain, of the variables x1, x2, x3 based a set of
membership function D11, D12 and D13. The inner-product Emor (3,5,8)
space for the fuzzy set Di , which defines the fuzzy region in
Mor (5,8,11)
the antecedent space, can be described as:
Fnoon (8,11,14)
Anoon (11,14,17)
(3)
Even (14,17,20)
Thus, the degree of fulfillment is represented as:
Dusk (17,20,23)
Night (20,24,23)

(4) Table 2. Parameters of membership function of output


variable
βi can be developed using other operator. The antecedent
part of the rule divides the input domain of x into a set of Very_low [0, 0, 40]
fuzzy sub-domains. Each domain is related to a fuzzy set. The
number of rules M can be given as: Low [0,40,100]
Medium [40,100,150]
(5) Moderate [100 150 200]
Where n is the size of the input and Termi is the number of high [150,200,250]
terms in the ith antecedent variables.
Vhigh [200,250,300]
3.2 Proposed Model V2high [250 300 350]
According to the characteristics of traffic flow and traffic
volume, Time of the day and day of the week can be used as V3high [300 350 400]
input variables of traffic prediction model system. The
output variable is the traffic volume. Traffic volume 3.2.2 Fuzzy Control Rules
prediction model and its fuzzy inference model’s block Once the Fuzzy membership functions are defined for input
diagram are shown in fig 1. and output variables, next step is to construct the fuzzy rules
for processing the fuzzy output. In the proposed model, rules
3.2.1 Fuzzification are
To fuzzify input and output domain, the number and
quantification levels of a variables have to be estimated. In
this paper, fuzzy statistics theory and traffic volume data for

2
International Journal of Computer Applications (0975 – 8887)
Volume 107 – No 17, December 2014

Table3.Fuzzy rules for the implementation of the proposed model

Midnit Dawn Emor Mor Fnoon Anoon Even Dusk Night


Mon Low Very_low Low High High V2high high Medium
Tue medium low low moderate High Very_high high moderate Medium
Wed moderate low low moderate High Very_high high Moderate Moderate
Thur medium low Low modeate High Very_high high Moderate Medium
Fri medium low low High V2high V3high V3high Very_high Medium

Fuzzy Inference

Day Fuzzy Rules

Traffic Volume
Defuzzification
fuzzification

week

Output
Input

Time of day Inference and Calculation


Fig 1. Traffic volume prediction system & fuzzy inference model

Rules reliability

Fig 1. Traffic volume prediction system & fuzzy inference model

Fig 2a. Membership function of Day of the week

Fig: 2b. Membership function of Time interval of the day

3
International Journal of Computer Applications (0975 – 8887)
Volume 107 – No 17, December 2014

mostly developed based on the data collected from Delhi-hardwar highway (NH-58 ) for week days. After that average data for each
days are calculated on the hourly basis .In this paper, fuzzy control rules are developed based on the hourly traffic volumes data.After
reviewing the data, a total of 44 rules are obtained. If the antecedences of several rules are the same, then the rule, which has the
largest weight, is kept. Rules designed for the proposed system are shown in Table 3.

3.2.3 Defuzzification
According to the input variables, their member functions and 41 effective control rules. The traffic volume at particular time and day
can be obtained by defuzzification. In this paper the ‘CENTROID Method’ was used for defuzzification. This is the most widely used
method. This can be called as centre of gravity or centre of area method. It can be defined by

the algebraic expression:


z*= { ∫μc˷(z).zdz}/{ ∫ μc˷(z)dz} (6)
∫ is used for algebraic integration.
This approach is much closer to mass decision making
process in real life. The final fuzzy logic model for traffic volume prediction is shown in figure 4. Logic can be a potential method
dealing with structural and parametric uncertainties in the traffic Volume behaviour human expert knowledge expressed by linguistic
variables is a powerful tool enabling fuzzy models to deal with uncertainties and inaccuracies. Membership functions, fuzzy control
rules of fuzzy logic to make traffic prediction model.

Fig 4. Traffic Volume Prediction Model

4
International Journal of Computer Applications (0975 – 8887)
Volume 107 – No 17, December 2014

4. RESULTS AND DISCUSSION [3] Zheng, W., Lee, D. H. and Shi, Q.2006. Short-term
After applying the available test data into the proposed traffic freeway traffic prediction: Bayesian combined neural
volume prediction model traffic volume will be predicted . network approach, Journal of Transportation Engineering
For testing the accuracy of the model prediction, two error , 114–121.
formulas are used, the first is used to test the accuracy of [4] Cheslow, M., Hatcher, S. G. and Patel, V. M.1992. An
single sample and second is used to the accuracy of the whole initial evaluation of alternative intelligent vehicle
sample. The difference between predicted traffic volume highway systems architecture. Rep. No. 92w0000063,
corresponding to the observed traffic volume of week days is MITRE Corporation, Bedford, Mass (1992).
shown in Fig 5 . The total error in the proposed model is
calculated by the Mean Absolute Percentage Error [5] Topuz , V.2007.Traffic Demand Prediction Using ANN
[35](MAPE).With the proposed method on average the Simulator, In: KES 2007/WIRN 2007, Part I, LNAI
MAPE error for week days is 15% The error shows that the 4692, Springer-Verlag, Berlin Heidelberg, 864-870 .
predicted traffic volume have relatively good correlation with [6] Okutani, I. and Stephanedes, Y. J.1984..Dynamic
the actual traffic volume.This reduce the MAPE error by prediction of traffic volume through kalman filtering
modifying the membership functions and rules.However, for theory, Transp. Res., Part B: Methodol, 18(1), 1–11.
some points, as shown in the fig 5, there are still have large
error and this is caused by the burstiness of traffic condition. [7] Xie, Y., Zhang, Y. and Ye, Z.2007. Short-term traffic
volume forecasting using Kalman filter with discrete
wavelet decomposition. Computer-Aided Civil and
Infrastructure Engineering, 22 326–334.
[8] Ross, P.1982.Exponential filtering of traffic data.
Transportation Research Record 869, 43–49.
[9] Smith, B. L., and Demetsky, M. J.1994. Short-Term
Traffic Flow Prediction: Neural Network Approach, In
Transportation Research Record 1453, TRB, National
Research Council, Washington, D.C., 98–104.
[10] Van Lint, J. W. C., Hoogendoorn , S. P., and Van Zuylen
, H. J, Accurate Freeway Travel Time Prediction with
State-Space Neural Networks Under Missing Data.2005.
Transportation Research Part C, Vol. 13, No. 5-6 347–
Fig 5. Simulated and Predicted Volume of week days 369.
5. CONCLUSION [11] Vlahogianni, E. I., Karlaftis, M. G. and Golias, J.
Modelling a traffic volume prediction model has proven to be C.2005.Optimized and Meta-Optimized Neural Networks
a very difficult task .The highly nonlinear nature of traffic for Short-Term Traffic Flow Prediction: A Genetic
behavior necessitates the use of fuzzy logic for the modeling Approach, Transportation Research Part C, Vol. 13, No.
and development of traffic volume prediction model. Fuzzy 3, 211–234.
logic can be a potential method dealing with structural and
parametric uncertainties in the traffic Volume behavior human [12] Park, D., and Rilett , L. R.1999. Forecasting Freeway
expert knowledge expressed by linguistic variables is a Link Travel Times with a Multilayer Feedforward Neural
powerful tool enabling fuzzy models to deal with uncertainties Network, Computer-Aided Civil and Infrastructure
and inaccuracies. In the current study membership functions, Engineering, Vol. 14, No. 5, 357–367.
fuzzy control rules of fuzzy logic is used to make traffic [13] Yin, H. B., Wong , S. C., Xu, J. M., and Wong , C.
prediction model. The above model has limitation that fuzzy K..2002.Urban Traffic Flow Prediction Using a Fuzzy-
control rules were designed on small data set .The result Neural Approach. Transportation Research Part C, Vol.
shows this approach is satisfactory and wide applicability in 10, No. 2, 85–98.
the area of Intelligent Transport System. However, there are
still gaps to do improve the reliability and efficiency of the [14] Li , L., Lin ,W.H. and Liu ,H.2006. Type-2 fuzzy logic
fuzzy logic model. In addition, new extensive data sets will approach for short-term traffic forecasting. Intelligent
be necessary to provide the basis for constructing a high Transport Systems, IEE Proceedings, 153(1), 33 – 40.
quality model. [15] Su, H., Yu, S. 2007. Hybrid GA Based Online Support
Using extensive data sets MAPE error can be minimized using Vector Machine Model for Short- Term Traffic Flow
neural network along with fuzzy . Forecasting In: APPT LNCS 4847, Springer-Verlag
Berlin Heidelberg,743-752.
6. REFERENCES [16] CHEN, B.P. and MA, Z. 2009. Short-term Traffic Flow
[1] Kajackas, A., Vindašius and Š., Stanaitis.2009., Prediction Based on ANFIS, International Conference on
Vehicle Communication: Emergency Message Delay Communication Software and Networks ICCSN, ISBN
Distribution , Electronic and Electrical Engineering. – 978-0-7695-3522-7, IEEE, 791-793.
Kaunas: Technologija , 8(96), 33–38.
[17] SADEK, A, W. 2007. Artificial Intelligence Applications
[2] Zeng, in Transportation, In: Transportation Research E-C113.
D., Xu , J., Gu, J., Liu, L. and Xu, G.2008. Short term Artificial Intelligence in Transportation Information for
traffic flow prediction using hybrid ARIMA and ANN Application. Transportation Research Board, ISSN 0097-
models, Workshop on Power Electronics and Intelligent 8515, 1-6 .
Transportation System, 621-625.

5
International Journal of Computer Applications (0975 – 8887)
Volume 107 – No 17, December 2014

[18] Smith, B.L.and Demetsky, M.J.1994. Short-term Traffic [28] Lotfi ,A. and Garibaldi ,J. M.2004. Applications and
Flow Prediction: Neural Network Approach.Transp. Res. Science in Soft Computing. Springer, ISBN: 3-540-
Rec. 1453, 98–104 . 40856-8.
[19] Celikoglu, H.B and Cigizoglu, H. K .2007, Modelling [29] Kosko ,B. 1998.Fuzzy systems as universal
public transport trips by radial basis function neural approximators, in Proceedings of Int. Conf. Fuzzy Syst.,
Networks In: Mathematical and Computer Modelling , 1153–1162.
Elsevier 45 ,480–489.
[30] Babu˜ska ,R.1996 Fuzzy Modeling and Identification.
[20] ] Ruimin Li, R , Hua ,p. u. , Lu, H. 2009. Combined PhD thesis, Delft University of Technology.
Neural Network Approach for Short-Term Urban
Freeway Traffic Flow Prediction, ISNN 2009, Part III, [31] Babu˜ska ,H. A., . Braake ,R., Krijgsman , A. J. and
LNCS 5553, Springer-Verlag Berlin Heidelberg 2009, Verbruggen . H. B.1996. Comparison of intelligent
1017–1025. control schemes for real-time pressure control, Control
Engineering Practice, vol. 4, 1585–1592.
[21] ]Abdulhai, B., Himanshu, P.and Will, R.2002.Short-
Term Traffic Flow Prediction Using Neuro-Genetic [32] Sheta , A. 2005.Modeling the Tennessee Eastman
Algorithms, Journal of Intelligent Transportation Chemical Reactor Using Fuzzy Logic. New York, USA:
Systems, 7: 1, 3 —41. Book Chapter. The ISE Book Series on Fuzzy System
Engineering-Theory and Practice, Nova Science, ISBN:
[22] Wang , L. X. and Mendel ,J. M.1993. Fuzzy adaptive 3-540-25322-X.
filter: with application to nonlinear channel equalization,
IEEE Trans. Fuzzy Syst., vol. 1,no. 3, 161–170 [33] Sheta ,A., . Oznergiz ,E., Abdelrahman, M. and Babuska,
R.2009.Modeling of hot rolling industrial process using
[23] Brown ,M. and Harris ,C. J.1994. Neurofuzzy Adaptive fuzzy logic, in Proceedings of the ISCA 22nd
Modeling and Control. Hemel Hempstead: Prentice Hall International Conference on Computer Applications in
Industry and Engineering (CAINE-2009), San Francisco,
[24] Wang, L. X.1992. Fuzzy systems are universal CA, USA, 81–86
approximators, in Proceedings of IEE Int. Conf. on
Fuzzy Systems, San Diego, USA, 1163–1170. [34] Abdelrahman, M., Sheta ,A., Deabes ,W.2009. Fuzzy
mathematical modeling for reconstructing images in ECT
[25] Dubois ,D. and Prade, H.1992. Fuzzy sets in approximate of manufacturing processes, in Proceedings of the
reasoning: part 1, Fuzzy Sets and Systems, vol. 40,143– International Conference on Computer Engineering and
202 Systems (ICCES2009), Ain Shams University, Cairo,
[26] Yager , R. R. and Filev, D. P. 1994. Essentials of Fuzzy Egypt.
Modeling and Control. John Wiley, New York.. [35] Taylor, J. R.1982. An introduction to error analysis: the
[27] Lotfi ,A.2002. Application of Learning Fuzzy Inference study of uncertainties in physical measurements,
Systems in Electricity Load Forecast. In report of the University Science Books, Mill Valley, CA.
EUNITE worldwide competitionon Electricity Load
Forecast using Intelligent Techniques (ISBN: 80- 89066-
41-0).

IJCATM : www.ijcaonline.org 6

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy