Printer Friendly

Research of customized shuttle bus route selection based on granular computing.

1. Introduction

Custom bus is a kind of exclusive public transport service, the service object is similar to the time of starting point, travel time, travel plans are relatively stable (mainly for commuter travel and travel), timeliness, comfort and high demand of group (Goncalves & Rocha, 2012; Xu, 2013). In recent years, the public on travel mode diversification demand rising, and the multiple factors promoting Internet plus "technology is widely used under the custom bus in a number of domestic city operation and gradually developed.

At present, the domestic bus companies are based on the reverse operation of passenger demand to operate custom bus. The basic process is: 1.first by the passenger to the bus companies to submit their own needs; 2.bus companies combined with the needs of passengers, the design of the best lines; 3. carry out the recruitment of passengers; operation. This line operation mode can effectively save the cost of public transport enterprises in the stage of investigation of demand, but there are obvious shortcomings: one is the demand for access to information is difficult, it is difficult to provide decision support for the operation of large data line. Since the line opened by the passengers need to submit travel demand, so the public transport enterprises, access to demand information is passive, which will cause the number of effective demand information can get less, low efficiency, directly affect the correctness of the line opened decision; two is the influence of the experience of passenger travel, resulting in the loss of customers. According to the literature (WANG, 2015), Beijing city bus ride customized passengers, there are about 60% before driving to work, this part of the main line of custom bus passengers booking accessibility, convenience, comfort is more sensitive, so in the case prior to the travel demand, if the demand is not met, it is easy to cause this some of the passengers not try to use custom bus travel. Thus, the custom bus lines operating plan is reasonable, to a large extent depends on the accuracy of the demand for investigation, and the current passive survey methods are difficult to do.

With the increasing of traffic data collection methods, the spatial and temporal data of traffic travel is increasing. Through the analysis of traffic data mining on these massive, multi-source, heterogeneous, can be extracted the travel characteristics of city population, so as to provide a guarantee for the scientific development of public transportation enterprise custom bus lines operation scheme. Granular computing (Computing Granular) is a new concept and computing paradigm of information processing, which covers all theories, methods, techniques and tools of all relevant granularity. The theory, fuzzy set theory, rough set theory, a superset of the theory of quotient space and interval calculation, a branch of soft computing science (LI, 2005). Granular computing is not limited to the details in the process of information processing, so it is a unique advantage in the analysis of massive, fuzzy, uncertain and incomplete information. The size calculation of the travel spatio-temporal data mining analysis of city public traffic, commuter travel characteristics (ZHANG, 2007), including commuting time and location, direction, route, preferences, and extraction may take a customized bus group number and distribution of information, provide the decision-making basis for the opening of custom bus lines.

2. Granular computing theory

Granular computing theory to the "size" as guiding ideology, the core idea is that everything can be in different dimensions, different levels of definition, measurement and reasoning, thus forming different integrated particles. Particle size calculation process is based on the actual needs, to find the problem to study the optimal dimension and level, so as to ensure that the problem can be quickly and accurately obtained the approximate solution of the problem (ZHANG, 2007).

A simple case of granularity thought: a group of some existing influenza patients with clinical data D, data attributes as { M | gender, age of onset, date, whether fever...}, In order to extract the information from the data, it can be used to measure the data from different levels of granularity. To obtain the influenza virus infection in the existence of gender differences, gender, the male and female {} for particle size, particle size of data integration, this level is relatively rough; for the influenza virus infection population information needs to be further fine granularity, the gender of {male and female age} [intersection] {age|< 10, 11~20, 20~30, 30~40, 40~50, 50~60, >60} as the particle size, particle of data integration, in order to get what one age, what kind of people belonging to high-risk groups.

1. Information granule

To the final domain (or can obtain information space) [O.sub.r], with no resolution or similarity of the relationship [P.sub.r] between the division [O.sub.r], each of which is known as the information particles [X.sub.r], which [P.sub.r1] [??] [P.sub.r2]. U can be one of the methods of clustering, coupling, dividing and so on. The tools used are fuzzy set, rough set, quotient space theory, etc.

2. Information granularity

The amount of information [X.sub.r] contained in the information granule is called the information granularity. If the domain [O.sub.r] is discrete, the information granularity represents the number of data contained [X.sub.r] in the information granule; if the domain is continuous, it represents the measure of the length of the information particle [X.sub.r]. Information granularity can be used to describe the coarse and fine, when dividing the relationship [P.sub.r1] [??] [P.sub.r2], then to divide the particles [P.sub.r1] than the coarse grain [P.sub.r2].

3. Information extraction

Carried out mining operations on the information particle "X"_"R", the original data will be no intuitive meaning processing process can obtain the information of the objective.

3. Line selection model

The basic principle of customized bus route optimization model based on granular computing is the traveler spatio-temporal data is hierarchical clustering using fuzzy quotient space theory, the choice of travel characteristics (demand, direction, tools etc.) homogeneous target groups, and to target population size as system bus lines provide the basis for the optimization (YAO, 2010). First of all, the establishment of travel mode choice factors evaluation system, and carries on the attribute reduction, improve the efficiency of the next cluster; secondly, construct the homogeneous travel fuzzy similarity matrix, and calculate the dynamic cluster analysis; finally, according to the operation strategy of bus companies choose different strategies, adjusted in different level of granularity and get the final clustering result.

The basic steps to build a customized bus route optimization model based on granular computing are as follows:

Step 1: Construction of travel mode selection impact factor evaluation system

From three aspects of traffic characteristics, individual characteristics and trip characteristics, the construction of 3 first level indicators, 11 two level indicators. The traffic features include traffic policy, traffic supply and traffic distribution; individual characteristics including gender, age, occupation, income, education degree; travel characteristics includes travel distance, travel time, travel destination (ZHANG, 2012), as shown in figure 1.

Step 2: determine the level of granularity of each evaluation factor, and reduce the evaluation factor.

Calculate each travel mode influence factor sensitivity using rough set, and according to the importance of the sort; according to the ranking results, first from fine-grained to multi-dimensional travel mode choice factors set of metrics, and according to the size of data on the impact factor of the grain coarsening gradually, attribute delete some weak correlation, improve the calculation efficiency of subsequent clustering.

Step 3: the collection of the impact factor data for Standardization Make [[alpha]] the value of the first influence factor i contained in the l object. For each travel mode to select the corresponding factors of all the data mean [[bar.[alpha]].sub.l] = 1/n [n.summation over (i=1)] [[alpha]] and variance [S.sub.l] = 1/n [n.summation over (i=1)][([[alpha]] - [[bar.[alpha]].sub.l]).sup.2], and then all the data into a standard value [[alpha].sup.*] = [[alpha].sup.*] - [[alpha].sup.*.sub.lmin]/ [[alpha].sup.*.sub.lmax] - [[alpha].sup.*.sub.lmin].

After the first standardized treatment, the value [[alpha].sup.*] = [[alpha].sup.*] - [[alpha].sup.*.sub.lmin]/[[alpha].sup.*.sub.lmax] - [[alpha].sup.*.sub.lmin] is not determined in the [0,1] range, which needs to be processed by two times, [a.sup.*.sub.lmax] and [a.sup.*.sub.lmin]

are the maximum and the minimum value [[[alpha].sup.*.sub.1l], [[alpha].sup.*.sub.2l] ,***, [[alpha].sup.*]].

Step 4: construct the fuzzy similarity matrix, and cluster analysis on the standardized sample data.

Defined [Q.sub.ij] [member of] [0,1 ] as the similarity between any two objects [[alpha].sub.i] and [[alpha].sub.j] the value of the degree of representation, calculation:

[Q.sub.ij] = 1 - M [[P.sub.1] [absolute value of [[alpha].sub.i1] -[[alpha].sup.*.sub.j1]] + [p.sub.2][absolute value of [[alpha].sub.i2] - [[alpha].sub.j2]] + *** + [p.sub.n][absolute value of [[alpha]] - [[alpha].sub.jn]]]

Construct similarity matrix. [p.sub.n] ensure that the value of the weight , M is in the interval [0,1], and the clustering method is used to get the clustering results.

Step 5: determine the threshold [phi], and adjust the size of the conversion, until the clustering results are satisfied.

According to the solution requirements and the corresponding velocity, the threshold value [phi]) is determined, and the satisfactory solution is obtained at different levels of grain size.

4. Model validation

4.1. Data preparation

Select the influencing factors to the users ultimately choose sensitive customized bus travel for quantitative analysis of the way to travel, 542 sets of data collected by the questionnaire; in addition, collected in Yantai from July 20, 2015 to July 26th, the Yantai public transportation card data bus APP data, a car and driver bayonet data (the driver data automatic matching by the vehicle data). The bus card and APP data is mainly used for screening by bus travel to customized bus travel crowd; a car and driver bayonet data is mainly used for screening by private car travel to custom bus travel crowd.

4.2. Attribute reduction

Establish travel mode to select the influence factors of the discretization of the rule, the collected sample questionnaire survey data to discrete expression, to facilitate the next step of attribute reduction, as shown in table 1:

Table 1--The rules of discrete data

                    Traffic characteristics

Discrete     policy     supply    Control    Gender

1             Yes        Yes        Yes       Male
2           commonly   commonly   commonly   female
3              no         no         no

                      Individual characteristics

Discrete     Age     Occupation     income        education
value                             (thousands)

1           18-35      fixed         <1.8         Graduate
2           36-50       free        1.8-3.2      University
3           51-64                  3.2-5.5      high school
4            >65                     >5.5       Junior school

              Trip characteristics

Discrete    Distance   Time     Cost
value         (km)     (min)   (yuan)

1              <3       <10      <5
2             3-5      10-30    5-10
3             5-10     30-50   10-20
4             >10       >50     >20

The investigation on traffic characteristics, is in Yantai city traffic policy, traffic supply and traffic control to the respondents whether there is influence of travel mode choice, options include effect, general effect, no effect three; individual characteristics of the project and the characteristics of the travel income division standard, is based on the actual situation of the local Yantai to determine.

According to the discrete rules in Table 1, the discrete expression of the impact factor sampling survey data is obtained, and the decision table is constructed. Based on the resolution matrix heuristic algorithm, the formula (1) is used to calculate the sensitivity of each primary trip mode to the choice of travel mode:

f(c) = [n.summation over (i=1)] [n.summation over (l=1)] [[alpha]]/[absolute value of []] (1)

The number of data contained in each of the factors that affect the. The sensitivity of each influence factor is calculated by the calculation, and the results are sorted according to the size, as shown in table 2:

Table 2--Sensitivity of Influential factor and scheduling

Properties          policy   supply      Control

Importance degree   2.0561   1.9324      2.5769
Properties          income   education   distance
Importance degree   3.1836   2.8231      2.1459

Properties          Gender             Age      Occupation

Importance degree   3.0783             3.2315   3.1821
Properties          Time consumption   Cost
Importance degree   2.7637             2.9322

According to the influence factor on the way to travel from sensitive sort, followed by the traffic supply and traffic policy, traffic control, travel distance, travel time and travel cost of travellers, level of education, gender, occupation, traveller income, traveller age. The surveyed sample data analysis based on a custom bus as a traffic information technology based on the convenient and comfortable public transportation, the service object is affected by age, gender, occupation, income, cost, time and other factors, the literature mentioned 60% customized bus tourists from the driving population conclusions are basically consistent. This is because the private car travel to customized bus travel crowd more and better on the travel time, travel cost affordability. In addition, we can see that the young and have a fixed occupation, income groups prefer the custom bus; whether you choose Custom bus travel and transportation supply, policy and travel distance and other factors associated with weak, analysts believe that this has a certain relationship with the Yantai city road network and traffic circle curing range is small.

Select the main traffic distribution in Yantai city of Wanda Plaza to the Sports Park as a custom bus line terminal, the maximum value of the clustering analysis of the survey data and the data bus card, APP bus data, with traffic matching to path distance discriminant rules for standard structure size, according to the literature the standardized data value [phi] = 0.82 as a threshold, to draw 3 lines, as shown in figure 2. Among them, a path (via South Street--two street--Guanhai Road) bear the full flow of 85%, green label; path two (the South Street--Victory Road--Red Road--sea view) bear the full flow of 78% red label; path three (via South Street--Jiefang Road--South--Ying Xiang Road Mountain Road Hong Kong--East Street) bear the full flow of 52% yellow mark. Therefore, we recommended route 1 for custom bus lines in operation, on his way to a commercial center in Luming station.

5. Conclusion

Build a custom bus route optimization model theory in dealing with massive, heterogeneous data traffic advantage is calculated by particle size, this model uses the original data preprocessing of rough set, excluding invalid and redundant data attributes, thus greatly reducing the amount of calculation; at the same time by using cluster analysis, clustering of data collection, the custom bus lines the traffic is relatively concentrated. However, as limited by the means of data collection, the amount of data and the number of less, in the future to consider adding more users attributes to improve the accuracy of decision making.

Recebido/Submission: 03/05/2016

Aceitacao/Acceptance: 23/08/2016


This paper is supported by Yantai science and technology project: "Research and development of customized public traffic information service system based on traffic time and space big data mining" (2015YT06000212).


Goncalves, J. J., & Rocha, A. M. (2012). A decision support system for quality of life in head and neck oncology patients. Head & Neck Oncology, 4(1), 1.

LI, D., MIAO, D., ZHANG, D. (2005). An Overview of Granular Computing. COMPUTER SCIENCE, 32(9), 1-12.

Mora, A. D., & Fonseca, J. M. (2014). Metodologia para a detecao de artefactos luminosos em imagens de retinografia com aplicacao em rastreio oftalmologico. RISTI--Revista Iberica de Sistemas e Tecnologias de Informacao, (13), 51-63.

WANG, Z., WANG, R. (2015). Adaptive Management of Beijing Urban Traffic: A Case Study of the Customized Transit Bus Service. Modern Urban Research, 3, 2-8.

Xu, K., Li, J., Feng, J. (2013). Discussion on Subscription Bus Service. Urban Transportation of China, 11(5), 24-27.

YAO, C., LUO, X., Henk, V. (2010). Short-term Traffic Flow Forecasting Based on Coupling of Rough Set and Neural Network. Journal of Highway and Transportation Research and Development, 27(11), 104-107.

ZHANG, R., YANG, J., LEI, X. (2012). Random Elasticty Analysis on Urban Travel Mode Choice. Journal of Transpotation Systems Engineering and Information Technology, 12(2), 132-136.

ZHANG, K., WANG, X.g, L, H. (2007). Transport Information Granular Computing Introduction Technical Architecture and Development Strategy. Journal of Highway and Transportation Research and Development, 24, 134-139.

Chen Yao,

Ludong University, Yantai 264000, China
COPYRIGHT 2016 AISTI (Iberian Association for Information Systems and Technologies)
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2016 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Yao, Chen
Publication:RISTI (Revista Iberica de Sistemas e Tecnologias de Informacao)
Date:Oct 1, 2016
Previous Article:Research on hospital human resource management based on cloud platform.
Next Article:Research on performance and influence factors of industrial technology innovation strategic alliances based on cloud computing data mining.

Terms of use | Privacy policy | Copyright © 2022 Farlex, Inc. | Feedback | For webmasters |