Академический Документы
Профессиональный Документы
Культура Документы
Abstract
1. Introduction
The main stakeholders in the transport sector are as follows (see Figure 1)
Transport authorities
and
and have
regarding
paysrepresentatives
responsability for
Elects their representatives
servicdelivered
transportation
that is delivered
Accounts regarding the
pays taxes
Organizefor
taxes
transportation
service Accounts
responsibility
ts their
and
D
Operators Users
The providers and operators of the services are transport companies, drivers
associations, and owners associations (Sohail et al., 2004). The operating companies
of transport networks will manage the public transport service when the transport
authority gives them the charge to do so under an agreement that specifies the
responsibilities and sets targets for proposals and revenues. These are regularly
evaluated and revised.
The traveler, the passenger or more accurately the user of public transport, is the one
who co-produces and finances the transport company. All users who use public
transportation are required to submit their transport subscription whatsoever
(subscription to lines network, subscription to school transport, etc...). The users are
particularly low-income households including men, women, children, elderly, and
disabled members of the community (Sohail et al., 2004).
Toni et al (2013) take into primary consideration the users' expectations. The users '
point of view is linked to their perceived quality of the delivered service and can be
considered the main driver of the investment choices to improve the service quality.
Several quality dimensions have been proposed to evaluate the quality of service
(Leurent 2009). These include punctuality, comfort, cleanliness, passenger
information etc.
The transport authorities can add other criteria that relate to their concerns
such as social responsibility, safety and security (Hsing et al., 2000). According to
Toni et al (2013), the transport authorities’ point of view is affected by matters
relating to equipment, in terms of quantity, quality and safety, and the environmental
impact of the service.
The operators’ point of view (particularly Transport Company) essentially
tends to focus on costs efficiency / effectiveness and service-effectiveness (Bertini et
al., 2003).
The transport performance evaluation should necessarily reflect the
expectations of different stakeholders like the operators and the transport authorities
that are directly or indirectly involved in the transit service (Toni et al, 2013).
Consequently the key performance metrics on which the performances of transport
companies are evaluated with respect to how much they are meeting stakeholder
expectations and improving the attractiveness of public transport are:
For public transport users, service quality is fundamental. Improving the quality of
service be at the heart of transport policy makers (AFNOR, 1997).
Annals of Management Science 69
Table 1. Dimensions of service quality (Source adapted from Eboli and Mazzula .,
2007)
Dimensions Description
Punctuality Reliability of the scheduled runs, Punctuality of the runs;
Comfort Bus crowding Comfort of bus seats ,
Availability of shelter and beaches at bus stops
Cleanliness Cleanliness of bus interior, seats and windows,
Cleanliness of bus exterior
Passenger Information Availability of schedule/maps on bus ,Availability of
schedule/maps at bus stops, availability of information by
phone internet ,Personnel appearance ,Personnel
helpfulness
Environmental impact Air condition on bus, Level of vibration on bus , air
pollution, the greenhouse effect, the smells, the
congestion
Several attributes of safety and security are identified by Eboli & Mazzula (2007)
such bus reliability, competence of drivers, security against crime on bus, security
against crime at bus stops , number of accidents etc .
Indeed transport companies must give great importance to the safety of
passengers and staff on public transport. Security represents a crucial issue for
public transport (Labathe, 2005).
Cost-
efficiency
Service Outputs:
Service Inputs:
Vehicle Hours,
Labor, Capital, Fuel Vehicle Kilometers,
Capacity Kilometers
Operationnel
performance
Service
Consumptions:
Cost Passengers, Passenger Service
effectiveness effectiveness
Kilometers, Operating
Revenue
Figure 2. Relationship between efficiency and effectiveness indicators (source
adapted from Bertini & El Geneidy, 2003)
In decision analysis, the choice of consistent family of criteria presents the most
important step required by multi-criteria approach (Roy, 1985).
Annals of Management Science 71
This is also considered an important and objective element in assessing the quality
of service. Indeed the respect for time schedule improves transport service quality.
The more respect that transport operators have for time schedule the more punctual
that vehicles arrive at arrival point and the more satisfied users feel. This criterion is
72 I. Zaouali and H. Khouaja
This criterion measures the degree of cleanliness of stations and the inside of the
vehicles provided by the transport company. This criterion is to be maximized. It is
measured with qualitative scale.
This assesses psychological comfort (feeling of safety (stations and on vehicles (not
flying) brightness) and physical comfort (sitting, standing). We use a qualitative
scale to measure this criterion.
A high quality and accessible user information and a warm welcome by the staff is
a prerequisite for a better and greater use of public transport. The lack of information
and respect for users then constitute a great impediment to the use of public
transport. We use a satisfaction scale to measure this criterion.
The pollution caused by passenger transport is noise and vibration, air pollution, the
greenhouse effect, the smells, the congestion (the concentration of automobile traffic
in a limited space, etc. This criterion evaluates the impact generated by vehicles on
users following the issuance of such nuisances.
This is an indicator of the level of road safety for the transport company. The
accident rate corresponds to the number of accidents observed during a given period
(year in this case) divided by the number of miles travelled by all vehicles in a year:
This measures the average age of the fleet of the company. We use as measuring
unit the number of years.
This indicator represents the relationship between the service consumption and the
service –inputs. The cost effectiveness is measured by the following rate:
This indicator represents the relationship between service –inputs and service-
outputs. The cost efficiency is measured by the following rate:
74 I. Zaouali and H. Khouaja
This indicator represents the relationship between the service -outputs and service-
consumption. The cost effectiveness is measured by the following rate:
Actions A
x xx Category 1
x
xx
Category 2
xx
xx
xx
Category 3
xx
ELECTRE TRI builds an outranking relation S using an index ∂(a, bh) (∂ (bh ,
a) ϵ [0,1], resp.) that defined the degree of credibility of the assertion aSbh (bhSa,
resp.), "a ϵA," h ϵB.
Exploitation of the relation S conducted to assign each alternative to a specific
category. Since the assignment of alternatives to categories does not result directly
from the relation S, an exploitation procedure is necessary. It requires the relation S
to be “defuzzy fied “ using a so called λ-cut the assertion aSbh (bhS a, resp.) is
valided if ∂(a, bh) ≥λ (∂(bh , a) ≥λ, resp.),λ being a `cutting level or the threshold
cut): a such that λϵ[0.5,1]. This λ-cut represents the preference situation between a
and bh:
∂(a, bh ) ≥ λ and ∂(bh, a) ≥ λ aSbh and bhSa aIbh , i.e., a is
indifferent to bh , (1)
∂(a, bh ) ≥ λ and (bh ,a) ˂ λ aSbh and not bhSa a›bh , i.e., a is preferred
to bh (weakly orstrongly),
(2)
∂(a, bh) ˂λ and ∂(bh , a) ≥ λ not aSbh and bhSa bh› a, i.e., bh is
preferred to a
Weakly or strongly
(3)
∂(a, bh )˂λ and ∂(bh, a) ˂ λ not aSbh and not bhSa aRbh, i.e., a is
incomparable to bh (4).
Annals of Management Science 77
Reference Preferential
Evaluation matrix parameters
Profiles
Outranking
relation
Assignment
procedure
Assignement of
alternatives
Our case decision-problem relates to regional transport companies. The input data
and parameter settings are presented in Table 7 below. Rows present the twelve
transport companies
(A1 to A12) and (wi, pi, qi) are respectively, the weight, the thresholds of
indifference and preference. Colons present a set of criteria (C1 to C2).The
intersection present he evaluation of each company on each criterion:
Sens Max Max Max Max Max Min Min Min Max Max Max
wi 0,10 0.10 0,07 0,07 0,08 0,05 0,14 0,07 0,10 0,10 0,12
qi 5 5 1 1 0 1 0 0 1 1 0,5
pi 10 10 2 2 1 2 0,1 0,5 3 3 1
Annals of Management Science 79
The use of the ELECTRE TRI method requires the arbitrary specification of a
number of parameters. First, a weighting system must be associated to the chosen
criteria in the first part (wi). Other parameters called thresholds are also defined.
Indeed , ELECTRE TRI allows us to take into account pseudo- criteria instead of
real - criteria by using thresholds of indifference (qi)and preference (pi) non- nulls.
(wi, pi, qi are respectively, the weight, the thresholds of indifference and
preference).
Four categories of companies are defined. These are:
1. The category of companies classified as”very successful” (VS). These are
high-performing companies in terms of quality of service, road safety and
economic efficiency and performance.
2. The category of companies classified as”successful” (S). These companies
have satisfactory performance in terms of service quality, road safety and
economic efficiency.
3. The category of companies classified as "less efficient" (LS). This category
of companies is characterized by poor performance in terms of service
quality, road safety and economic efficiency.
4. The category of "non-classified" companies (NC). These companies are
characterized by very poor performance in terms of service quality, road
safety and economic efficiency.
Each category is characterized by a reference profile (R3, R2, and R1). R3 (the
high profile) is greater than R2 (the satisfactory profile) and R2 is greater than R1
(low profile). Each profile represents the boundary between two adjacent categories.
They define the four categories already identified above (VS, S, LS, and NC). R3 is
the boundary between VS and S, R2 is the boundary between S and LS and R1 is the
boundary between LS and NC. Each profile depends on the performances of the
companies on each of the eleven criteria and preferences of different stakeholders
(importance of the criteria, etc.). R3 profile presents the vector of optimal values of
the criteria. R2 profile is the vector of satisfactory values. R1 profile is the vector of
lower values (see Table 7):
R3 90 90 5 10 5 0 0.3 6.5 32 36 5
R2 60 51 3 5 3 3 0.5 7,5 22 25 3
R1 30 30 0 0 0 5 0.7 9 12 18 2
From table 8 the reference profiles values are calculated taking into account
the stakeholders’ expectations and the judgments of some experts in the transport
field.
The weights are determined by a pair wise comparison of the criteria using
the approach developed by Saaty (1990). The results are then validated by the
relevant stakeholders intervening in the decision aid process (see table 8).
80 I. Zaouali and H. Khouaja
Weights 0.1029 0.1029 0.0443 0.0787 0.0848 0.0523 0.1386 0.067 0.103 0.103 0.1226
The weights are calculated using the analytic hierarchy process software. This
software also allows you to evaluate a posteriori the consistency of judgments using
the CI (Consistency Index). The purpose for calculating the consistency index is to
make sure that the original preference ratings were consistent. In practice, a CI of
0.1 or below is considered acceptable. Any higher value at any level indicate that
the judgments warrant re-examination (If the CI is greater than 0.1, the matrix
evaluations should be reassessed).
This method has several advantages. On the one hand it takes into account all
the tangible and intangible criteria, quantitative and qualitative. On the other hand it
is characterized by its simplicity in calculating the weights. It does not also
necessarily seek consensus but develops a representative compromise of various
judgments. Additionally, it allows us to evaluate the logical consistency of
judgments using the CI (Consistency Index).
As a reminder, we have three reference profiles such as R3> R2> R1 and therefore
four ordered categories VS > S > LS > NC. The pessimistic assignment procedure
does rankings as follows. The company Ei is compared successively with R3, R1
and R2 so that if:
Ei Outperforms R3, Ei is assigned to the VS category: The Company is
considered very successful
Ei outperforms R2 but does not outperform R3, Ei is assigned to the S
category: Ei will then be considered as successful.
Ei outperforms R1 but does not outperforms R2, Ei is assigned to the LS
category: The company is then considered less successful.
R1 outperforms Ei; Ei is then assigned to the NC category and the company
is assigned to the not- classified category.
The results of the pessimistic assignment procedure for Cut threshold λ=0.76
(the default value of the software) are:
From this table (table 10) we can conclude that only three alternatives are classified
to satisfactory performance. Others companies are assigned to poor performance.
Annals of Management Science 81
The results of optimistic assignment procedure for a cut threshold λ = 0.76 are:
VS : High performance
NC : non classified
From this table (table 11) we can conclude that more companies are classified
to satisfactory performance category.
It can be seen from table 10 and 11 that the sorting results of the ELECTRE
TRI method using two procedures are different. The pessimistic procedure tends to
classify the major companies to the poor performance category (LS). Only three
companies are classified to the satisfactory performance category (S). This
procedure is applied when a policy of prudence is necessary because it tends to
classify the companies to the lower possible category.
The optimistic procedure tends to classify more companies to the satisfactory
performance category (S). In fact this procedure tends to classify the companies to
the higher possible category in contrast to the pessimistic procedure.
We retain the results of the pessimistic procedure of assignment for sensitivity
analysis. The reason for this is that the optimistic procedure tends to classify the
companies to the higher possible category. This is in contrast to the pessimistic
procedure that tends to classify the companies to the lower possible category. This
procedure is applied when a policy of prudence is necessary. In ELECTRE TRI
method, the assignment depends primarily on the value of the cutting level and the
allocation procedure.
A sensitivity analysis with respect to thresholds cutting yields results that
assess the significance of changes in the allocation obtained. We distinguish three
types of changes, namely:
a) Change type 1: occurs when a company is moved from category C1 to C2, or
C2 to C 3 or C3 to C4 and vice versa.
b) Change type 2: occurs when a company is moved from category C1 to C3 or
C2 to C4 and vice versa.
c) Change type 3: occurs when a company is moved from category C1 to C4
and vice versa.
82 I. Zaouali and H. Khouaja
The results of the sensitivity analysis with respect to the variation of the
threshold cuts are summarized in Table 12.
The importance of changes for different thresholds cutting reveals that when
the cutting threshold varies between 0 .76 and 1, assignments remain the same.
There is no change type 2 or 3. There are two changes type 1 for λ = 0.6. This
indicates stability of the results.
The results are presented in the following graph (see Figure7).The company A1
outranks all the other companies. The second best position is occupied by A2 and
A4. Companies A7 and A11 are third. The fourth position is occupied by companies
A3 and A12 while A5 and A10 are ranked fifth. The company A8 occupies the 6th
place while the last class is occupied by A6 and A9.
We can conclude that company A1 is the most efficient. This ranking is stable
for different sets of weights considered. Company A4 and A1 rank first when road
safety criterion has a larger importance than those of service quality and operational
performance criteria. In the case where the quality of service criteria are considered
more important than the two other types of criteria, A1 is ranked higher than the
other companies in terms of overall performance. This result is consistent with those
obtained in absolute evaluation.
4. Summary
This study shows the applicability and usefulness of multi-criteria decision analysis
technique in evaluating the performance of transport companies. Absolute
evaluation poses problem in terms of sorting companies by categories when each
category has an intrinsic definition. That is, when each category has no reference to
other categories. Relative evaluation poses problems in terms of ranking of these
companies.
This paper was focused on the application of a multi- criteria method for the
evaluation of the performance of transport companies in Tunisia and for ranking
them according to their performance. To model this problem we began by building a
consistent family of criteria: 11 criteria grouped into three categories were used.
These categories are service quality, safety and operational performance in terms of
efficiency and effectiveness. We presented and applied firstly the ELECTRE TRI
method to assign the different alternatives in predefined categories. To test the
validity of the results, we did a sensitivity analyzes .These analyzes show that the
results obtained with the threshold method are stable. Then the ELECTRE III
method is applied to classify the companies.
The application of partial aggregation methods (ELECTRE TRI and
ELECTRE III), in evaluating the performance of public transport companies allows
us to have a global view. Its other advantages are:
a) The use of quantitative and qualitative data.
b) Possibility of using cardinal and ordinal criteria.
c) Limit the phenomena of compensation between the criteria.
d) Avoid the aggregation of multiple criteria in a single synthesis.
e) The possibility of using fuzzy or uncertain data by using the concept of
pseudo-criteria and thresholds.
References