PASSENGER TRAVEL CHOICE PREDICITON

BASED ON FUZZY LOGIC

Liu Mei, Li Jing

School of Economics and Management, Beijing Jiaotong University, System, Yuen Estate, Haidian District, Beijing, China

Kang Shu

School of Economics and Management, Beijing Jiaotong University, System, Yuen Estate, Haidian District, Beijing, China

Keywords: High-speed rail, Passenger travel choice, Fuzzy logic.

Abstract: Travelling choice about high-speed rail is very important to improve the level of management. The paper

put forward a prediction model based on fuzzy logic to predict the choices of the travelers , after compared

several methods. Then , combined with the survey data and the training and testing of the model , we

determined a effective model with fuzzy rules. The inspection findings of the model indicate that the choice

result predicated by the model based on fuzzy logic and the actual choice very close. Therefore, the

prediction method based on fuzzy logic is feasible. In addition, we put forward the concept of improving

factor, to provide reasonable proposals about how to improve the services and management quality of the

high-speed way.

1 INTRODUCTION

The issue about passenger travel choice is of

universal concern, the result of the prediction and its

reliability is very significant to improve the level of

management and make aid decision. As the most

effective way to solve the conveying of a large of

passengers quickly, High-speed rail has already

become the general development tendency.

Nevertheless, the studies on high-speed rail in our

country mostly focus on the field of technology,

while is lack in the studies on passenger travelling.

The essence of the travel choice decides its result

has strong randomness, and its impact factors have

obvious ambiguity and uncertainty. Therefore, how

to solve the problem caused by the ambiguity and

uncertainty is the difficult point of the prediction.

Although traditional soft computing methods are

strict , certain and accurate, they are not suit to solve

the practical problems under uncertain environment

in real life, such as the problem about passenger

travel choice. For this reason, the paper discussed

and put forward the prediction method for the travel

choice based on fuzzy logic.

Besides, we put forward how to distinguish the

most effective impact factors.

2 LITERATURE REVIEW

Fuzzy logic has been growing attention for engineers

since 1965 (Zadeh, 1965). In traffic field, traffic

signal control is an important application. However,

study on passengers travel is really in shortage. Most

study about traffic is about traffic signal control.

The first appearance of the fuzzy logic in traffic

signal control is due to Pappis and Mamdani( 1997).

About passenger travel, scholars’ study focused

on analyzing passengers’ demand. Qiang

Lixia and Yan Ying (2006) put forward the

differences in high-speed rail in china and other

countries. Go so far as to passenger travel choice,

most scholars analyzed it based on behaviouristics.

Chen Zhangming (2008) put forward passenger

travel choice by researching travelers’ action feature

.In brief, study on passenger travel choice based on

soft computer is such as never previously existed.

163

Mei L., Jing L. and Shu K..

PASSENGER TRAVEL CHOICE PREDICITON BASED ON FUZZY LOGIC.

DOI: 10.5220/0003477901630166

In Proceedings of the 13th International Conference on Enterprise Information Systems (ICEIS-2011), pages 163-166

ISBN: 978-989-8425-54-6

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

3 THE PREDICTION MODEL

BASED ON FUZZY LOGIC

3.1 Prediction System and its Model

Structure

In accordance with the expert’s suggestion about

passenger travel, we choose ticker price(TP),

departure time(DT),train environment (TE), train

security(TS), train speed(TSS) as the input variable

of the prediction system, the choice as the output of

the system. The fuzzy inference model structure is

shown in Figure 1, the key problem is to confirm

effective fuzzy rules.

Figure 1: Prediction system structure.

3.2 Fuzzification

Based on the survey data, we determined fuzzy sub-

sets and membership function for each parameter.

TP is expressed by{VC,C,M,E,VE},

and DT is expressed by{EL,VL,L,M,S,VS,ES},and

TE is expressed by {EL,VL,L,M,S,VS,ES}, and TS

is expressed by {EL, VL, L,M,S,VS,ES},And TSS is

expressed by {EL, VL, L, M, S, VS, ES}. The

choice is “yes” or “no”. The membership functions

of these parameters are determined shown in Figure

3.3

Fuzzy Rules

As for each effective record of the survey data, we

set up a relevant fuzzy rule. For example, to a

certain passenger, if he thought ticker price(TP) is

M, departure time(DT) is VS, train environment(TE)

is M, train safety(TS) is VS, train speed (TSS) is VL

,and the choice of his next travel is not choosing

high-speed rail, we can describe this rule with fuzzy

language as that

If (TP is M)and (DT is VS) and (TE is M) and

(TS is VS) and (TSS is VL) , then (choice in not).

(a) TP

(b) DT

(d) TS

(e) TSS

Output variable

Figure 2: Membership functions.

Accordingly, the weight of this rule is

This paper collected 1270 records about passenger

travel choice, then, divided to two parts equally. The

first part contained 447 records after deleting

reduplicate records, these were for the training of the



ICEIS 2011 - 13th International Conference on Enterprise Information Systems

164

fuzzy rules. The second part was used to examine

the efficiency of the prediction.

Therefore, we got 447 initial rules, we screened

these rules. If some rules’ premise were the same,

we retained the rule that its weight was biggest, and

deleted others. Then, we trained the rules with

simulink . In the end, we got 211 rules, as shown in

table 1.

3.4 Travel Choice Prediction Model

On the basis of decided input variable, their

membership functions, and 211 rules after trained

,we finally built the passenger travel choice

prediction model, shown in figure 3.

4 EXAM AND ANALYSIS MODEL

Make the second group data as the input of the fuzzy

logic travel choice prediction model, we got the

homologous prediction result. To exam the precision

of the prediction model, we used two error formula,

one was for confirming the prediction precision of a

single sample point, the other one was for

confirming the prediction precision of the whole

sample.

The exam result between the prediction result and

the actual passenger travel choice is shown in table

2.With the comparison between the prediction result

and the actual choice ,we got the accuracy rate of the

prediction model is 87%.

Figure 3: Fuzzy logic travel choice prediction model.

Table1: Fuzzy logic rules.

TP TT TE TS TSS CHIOCE

M S M S VL N

ES M M M M Y

M L S S VL N

VS M M S M N

…… …… …… …… …… ……

M M S S M Y

Table 2: comparison result.

Record number Actual choice Prediction result Comparison

1 Y Y same

2 Y Y same

3 Y Y same

4 Y Y same

5 Y N different

6 Y Y same

7 Y Y same

… … … …

447 N N same



PASSENGER TRAVEL CHOICE PREDICITON BASED ON FUZZY LOGIC

165

5 DISTINGUISH HIGHEST

IMPACT FACTOR

Effective fuzzy control rules not only provided the

relationship between input variable ant output

variable, they also reflected each impact factors’

effect on the final prediction result in a way.

Therefore, with this fuzzy logic model, we can exam

the most effective impact factor to move forward a

single step.

To a certain passenger, change the value of

each impact factor one by one. Then, get the

homologous prediction result with the model. After

that, we can compare the prediction result with the

actual passenger travel choice. The principle is that

after changing the factor, if the prediction result is in

the same with the actual choice, then, we can know

the degree of the influence of this factor.

6 CONCLUSIONS

It’s very difficult to build an accurate prediction

model to predict passengers travel choice. The

reason is that passenger’s choice about travel is

subject to a lot of factors, such as time, weather,

individual trait etc. These all factors can have huge

influence on passengers’ travel. Thus, this paper’s

ultimate aim is to put forward a feasible model to

predict passengers’ choices when they is going out.

Nevertheless, because of travel choice’s

randomness, discussing and getting a model to

predict passenger travel choice exactly is very

unpractical. The demonstration of this paper show

that to predict passenger travel choice with fuzzy

logic is feasible in a way. Of course, to improve the

reliability of this model, there are still a lot of

problems needing further study. In addition, a more

complete and accurate data base is also

indispensable, based on it, we can set up a model

with higher quality.

REFERENCES

Wu Qun -qi, Xu Xing. Mechanism research on travelling

choices of passengers. Journal of Chang'an University

(Social Science Edition) [N].2007,9(2) :13-16.

Chen Zhang-ming, Ji Xiao-feng. Study on Features of

Railway Passengers’ Travel Activities. Railway

Transport and Economy[J].2008,30(11): 23-25.

Gong Gu, Zhao Xiang-jun, Hao Guo-sheng, Chen Long-

gao. Study and Implementation of Genetic Algorithm

Based on Improvement of Search Space Partition.

Journal Of Henan University9natural Science [J]

2009, 39 (6) :631-636.UNIVERSITY9NATURAL

SCIENCE0 [J] 2009,39 (6) :631-636.

Huey-Ming Lee

．

Generalization of the group decision

making using fuzzy sets theory for evaluating the rate

of aggregative risk in software

development

．

Information Sciences,1999(113)

：

301-

311.

The Development of the world's high-speed railway.

Railway survey and design [j]. 2006,1: 54-56.

Zadeh, L. A., Fuzzy sets. Information and Control 1965,12

(2):94–102.

Voula C. Georgopoulos, Chrysostomos D. Stylios

Complementary case: based reasoning and competitive

fuzzy cognitive maps for advanced medical decisions.

Soft Comput [J] (2008)12

：

191-19.

．．．

A Zadeh Fuzzy Sets as a Basis for Theory of

Possibility. Fuzzy sets and Systems, 1978(1):3-28.

ICEIS 2011 - 13th International Conference on Enterprise Information Systems

166