• Editorial Board +
• For Contributors +
• Journal Search +
Journal Search Engine
ISSN : 1598-7248 (Print)
ISSN : 2234-6473 (Online)
Industrial Engineering & Management Systems Vol.19 No.2 pp.442-459
DOI : https://doi.org/10.7232/iems.2020.19.2.442

# A Combined Approach Based on Fuzzy SERVPERF and DEA for Measuring and Benchmarking the Quality of Urban Bus Transport Service at the Route Level

High School of Technology of Fez (HSTF), Sidi Mohamed Ben Abdellah University (SMBAU), Fez Morocco
*Corresponding Author, E-mail: karim.zehmed@usmba.ac.ma
December 10, 2019 March 13, 2020 May 8, 2020

## ABSTRACT

The most of previous studies evaluate the relative service quality at bus routes level regarding certain aspects without establishing any interaction between company and passenger’s point of view nor quantifying needed improvements. To overcome this gap, a three-stage approach based on the Fuzzy SERVPERF and Data Envelopment Analysis method has been adopted in this paper. The first stage includes an assessment of the perceived quality of each bus route, taking into account the SERVPERF model. The second stage involves using the DEA to calculate the perceived quality index from the SERVPERF dimensions scores. Both phases were conducted under a fuzzy environment to address the possible ambiguity and imprecision of the customer’s point of view in the assessment process. In the third stage, a quality effectiveness DEA model is proposed, in which the company and customers’ points of view are taken into account, respectively. Namely, the model identifies ineffective bus routes where the actual level of quality, as provided by company, could not be translated to highest possible levels of perceptions by customers, and provides potential improvements for these routes. To illustrate our approach, an empirical study is conducted about 25 bus routes in a major Moroccan city named Fez.

## 1. INTRODUCTION AND LITERATURE REVIEW

Improving the quality of urban public transport services is crucial for both community and transport operators. On the one side, a high level of quality is an incentive for people to use public transport instead of a private car. This shift may help in avoiding many issues such as traffic congestion, air and noise pollution, and energy consumption (Eboli and Mazzulla, 2008). On the other side, transport operators should improve the quality of service if they want to maintain or expand their market share and increase their financial profitability. Therefore, service providers should assess service quality regularly to ensure the effectiveness of the service.

The quality of urban public transport services could be assessed from two different points of view (Eboli and Mazzulla, 2011). The passenger’s point of view is re- presented by subjective measures that are qualitative and are based on the expectations/perceptions of customers (Eboli and Mazzulla, 2011). In this regard, various methodologies and techniques have been proposed in the last few decades that belong to three broader categories, including survey studies and interviews, statistical analysis of collected data, and multi-criteria decision making (MCDM) (Awasthi et al., 2011). The company’s point of view is represented by objective measures, that are quantitative and expressed as numerical values, which must be compared with fixed standards or previous performances (Eboli and Mazzulla, 2011). Some scholars have recently proposed a third perspective, which is focused on the combination of the two previous points of view (Zehmed and Jawab, 2019).

Most of the existing studies have been concentrated on evaluating the service quality of the entire system or comparing companies (Güner, 2018). However, a routelevel analysis can also provide essential insights for the improvement of the service quality. To this end, some recent studies have been dedicated to analyzing service quality at the route level using different methodologies and under different points of view. From the passenger’s perspective, Awasthi et al. (2011) present a hybrid approach based on SERVQUAL and Fuzzy TOPSIS for evaluating the service quality of Metro lines. Kanuganti et al. (2013) compare four different methods, namely numerical rating, fuzzy set theory, AHP, and fuzzy AHP, for calculating the level of service index of bus routes. Aydin et al. (2015) propose a framework combining statistical analysis, fuzzy AHP, fuzzy trapezoidal sets, and Choquet integral to evaluate the customer satisfaction of rail transit lines. Aydin (2017) propose method combines statistical analysis, fuzzy trapezoidal numbers, and TOPSIS to assess service quality levels of rail transit lines for multiple periods. Chica-Olmo et al. (2018) propose a combined method using nonlinear principal component analysis and a multilevel logit model in two steps to analyze customer satisfaction at bus lines level. Zehmed and Jawab (2020) propose a three-stage methodology combining importance-performance analysis and Data Envelopment Analysis to, jointly, identify service aspects and bus routes that should be prioritized for improvement actions. Finally, Nikel et al. (2020) developed a two-step cluster analysis to classify bus routes based on their operational characteristics, which is followed by a series of Importance-Performance models corresponding to each route type.

From the combined approach, the first attempt is made by Eboli and Mazzulla (2011), which analyze jointly subjective and objective quality measures and combined them in a single output indicator. Their approach has, however, been applied only on one bus route. Hassan et al. (2013) propose a framework based on TOPSIS to measure public transit service performance at the route level, taking into account the opinions of users, operators, and service providers. Barabino and Di Francesco (2016) developed a holistic framework called “TRANSQUAL,” deployed on three stages, namely characterization, measurement, and management of service quality. The framework is based on gaps analysis between desired/perceived quality and targeted/delivered quality, respectively. Finally, Güner (2018) proposes a two-stage MCDM approach. The first stage involves the use of AHP to determine the priority of each service quality attribute from the passengers’ point of view, and the second stage adopts TOPSIS to measure the delivered service quality and rank the bus transit routes.

The significant gap of all the above studies lies in the fact that they make only a parallel evaluation of the two points of view. Still, they do not establish any interaction between them in terms of effects and needed improvements. Indeed, the majority of studies evaluate the relative level of service quality of bus routes regarding certain aspects. However, they did not identify bus routes from where imitate best practice to improve quality or offer a quantification of needed improvements. In order to overcome these limits, the present paper proposes a threestage approach based on the Fuzzy SERVPERF and Data Envelopment Analysis method to measure and benchmark the quality of bus transport service at the route level. The first stage is an assessment of the perceived quality of each bus route, considering the SERVPERF model. The evaluation is based on fuzzy numbers to deal with the possible ambiguity and imprecision of the customer’s point of view in the assessment process. The second stage involves using the DEA method as an MCDM tool to aggregate SERVPERF dimensions scores into a single measure of perceived quality. In the third stage, a quality effectiveness DEA model is proposed where the company and customers’ points of view are taken into account, respectively. Namely, the model identifies ineffective bus routes where the actual level of quality, as provided by company, could not be translated to highest possible levels of perceptions by customers, and provides potential improvements for these routes.

DEA is a non-parametric technique for measuring the relative efficiency of a set of organizational units (Called “decision-making units (DMUs)”) able to transform multiple inputs (resources) into multiple outputs (services). DEA can take into consideration various inputs/ outputs as it does not require the specification of any particular functional form at the production frontier nor the weights attached to each input/output. DEA results identify the best efficient units as well as suggest the improvement that is required by all the other inefficient units to reach them. For these advantages, DEA has been used in this paper.

Using the DEA method to analyze the perceived quality and satisfaction is not new. It has been widely and effectively used in several sectors such as banking (Charles and Kumar, 2014;Khalili-Damghani et al., 2015), auto service repair (Lee and Kim, 2014), automobile industry (Lewis and Mazvancheryl, 2011), hotel industry (Najafi et al., 2015), public university (Mainardes et al., 2014), fertilizer brands (Mohanty and Senthil Kumar, 2017) and mobile phone brands (Bayraktar et al., 2012). However, this technique has never been applied to measuring the perceived quality in the context of bus transport. Furthermore, the DEA model where inputs and outputs quality indicators reflecting, respectively, the company and customers’ point of view is also not new in literature. Soteriou and Zenios (1999) and Adler and Berechman (2001) proposed DEA models to evaluate the quality of bank branches and airports, respectively. Nevertheless, this approach has not yet been applied in public transport at the route level. The only attempts for measuring the quality of bus routes using DEA, proposed by Lin et al. (2008), Sun et al. (2016) and (Tran et al., 2017), are based solely on objective measures of quality, which reflects the company’s point of view (Karim and Fouad, 2018).

The rest of this paper is organized as follows. Section 2 gives a brief overview of the methodological concepts used in this paper. Section 3 presents in detail the proposed methodology. In the fourth section, a case study is presented and results are discussed. Finally, section 5 provides conclusions.

## 2. PRELIMINARIES

### 2.1 SERVPERF Model

Various models were developed to measure the service quality from the customer’s point of view, but, traditional ones, namely (SERVQUAL = SERVICE QUALITY) and (SERVPREF = SERVICE PERFORMANCE), are still the most comprehensive and widespread used ones (de Oña et al., 2014). Parasuraman et al. (1988) first developed the (SERVQUAL), which is a survey instrument comprising 44 items for measuring both perceptions and expectations (22 for each one). Based on the gap model (Parasuraman et al., 1985), the service quality score in (SERVQUAL) is a function of the differences between customers’ expectations and perceptions. Although (SERVQUAL) gained popularity, it is criticized for its operationalization, validity, and about the fact of com-bining expectations and perceptions (Lee and Kim, 2014). As a way of addressing SERVQUAL’s shortcomings, Cronin and Taylor (1992) proposed the SERVPERF model in which the service quality score is measured based only on a customer’s perception through 22 items.

Both SERVQUAL/SERVPERF items are grouped into five dimensions, namely tangibles, reliability, responsiveness, assurance, and empathy. These dimensions are defined as follows (Parasuraman et al., 1988): Tangibles include physical facilities, equipment, and appearance of personnel. Reliability relates to the ability to perform the promised service dependably and accurately. Responsiveness is the willingness to help customers and provide prompt service. Assurance refers to the knowledge and courtesy of employees and their ability to inspire trust and confidence. Empathy refers to caring, individualized attention to customers. Besides, the service quality score could be measured at the item level, dimension level, or overall level. In the present paper, the measurement at the SERVPERF dimension level is considered.

### 2.2 Triangular Fuzzy Numbers

Customer satisfaction surveys in public transport commonly use questionnaires in which linguistic evaluation scales are employed to assess customer perceptions. Each linguistic scale is then associated with an exact numeric value. In particular, 5-point Likert scales, defined from “Not satisfied” to “Very satisfied, are the most widely adopted (de Oña, 2015). However, individuals perceive such linguistic expressions differently. Thus, merely summing up and averaging the scores of each respondent will result in a bias in service quality measurement (Yu et al., 2015). In response to this shortcoming, the fuzzy set theory (Zadeh, 1965) has been widely used to capture the respondent’s structure preferences and to assess the ambiguity of concepts associated with the respondent’s subjective judgment (Yu et al., 2015). In particular, linguistic terms can be better modeled by fuzzy numbers (Yeh et al., 2000). Accordingly, the fuzzy numbers have been recently employed, successfully, to evaluate the service quality and satisfaction in public transport services (e.g., de Aquino et al., 2018;Aydin et al., 2015;Aydin, 2017).

In this regard, triangular fuzzy numbers (TFN) are employed in the present paper to characterize the linguistic terms used by bus users to represent their judgment on perceived service quality. In particular, the membership of a TFN Ã is a function , which associates with each element in x in a real number in the interval [0, 1] (Zadeh, 1965). It’s defined as a set of equations (1), Where lA, mA, uA are real numbers and lAmAmA. The parameter mA corresponds to the maximum value of (i.e. 1 for normalized TFNs), whereas lA and uA are the lower and upper bounds of the definition interval respectively.

(1)

A TFN Ã can also be characterized by its confidence interval at the level α (α-cut):

(2)

Based on the extension principle suggested by (Zadeh, 1965), algebraic operations such as addition, subtraction, multiplication, and division of the triangular fuzzy numbers for $A ˜ = ( l 1 , m 1 , u 1 )$ and $B ˜ = ( l 2 , m 2 , u 2 )$ can be specified, respectively, as follows:

$A ˜ ⊕ B ˜ = ( l 1 + l 2 , m 1 + m 2 , u 1 + u 2 )$
(3)

$A ˜ ⊖ B ˜ = ( l 1 − l 2 , m 1 − m 2 , u 1 − u 2 )$
(4)

$A ˜ ⊗ B ˜ = ( l 1 × l 2 , m 1 × m 2 , u 1 × u 2 )$
(5)

$A ˜ ⊘ B ˜ = ( l 1 / l 2 , / m 1 / m 2 , / u 1 / u 2 )$
(6)

### 2.3 Data Envelopment Analysis

DEA is a non-parametric technique for measuring the relative efficiency of a set of organizational units (Called “decision-making units (DMUs)”) able to transform multiple inputs (resources) into multiple outputs (services). The relative efficiency of a DMU is measured by estimating the ratio of weighted outputs to weighted inputs and comparing it with those of other DMUs. DMUs that achieve a score of efficiency of unity (or 100%) are considered efficient, while DMUs with efficiency scores under 100% are inefficient. The first DEA model, proposed by Charnes et al. (1978), is called the CCR model and assumed constant returns to scale. Later, Banker et al. (1984) expanded the CCR model into the BCC model to take into account variable returns to scale situations. Additionally, a DEA model (CCR or BCC) can be oriented to inputs or outputs. The DEA model minimizes inputs for a specified output level within an inputoriented approach. Conversely, the DEA model maximizes outputs for a specified amount of inputs in an outputoriented approach.

In this paper, we deal with inputs/outputs that strictly reflect the quality indicators instead of inputs/outputs of the pure production process (transformation of physical resources into service quality). Thus, the use of DEA is oriented to analyze quality effectiveness rather than quality efficiency. Furthermore, the output-oriented CCR model is used rather than BCC one, as there is no strong theoretical background for using the BCC model. On the other hand, the CCR model provides, generally, better discrimination compared to BCC model for given numbers of inputs, outputs and DMUs (Podinovski and Thanassoulis, 2007). In this sense, Wojcik et al. (2019) argued that “the CCR models are the most rigorous ones, because they compare the respective DMU with all linear combinations of the other DMUs, and not only with their convex combinations as is the case with the BCC models (p, 563)”.

### 2.4 DEA as an MCDM Tool

Although the conventional goals of DEA and MCDM differ as MCDM techniques seek to prioritize a range of alternatives with contradictory criterion, several researchers tried to apply DEA as a technique of MCDM problems (Golany, 1988;Adolphson et al., 1990;Belton and Vickers, 1993;Doyle and Green, 1993;Stewart, 1996). Their rationale justifying this use is as follows: If inputs and outputs of DEA are substituted for attributions or evaluation criteria of MCDM and DMUs of DEA are considered as alternatives of MCDM, then the DEA problem accords with the MCDM problem (Adolphson et al., 1990). However, it would be pointless in MCDM to consider both inputs and outputs as measurement criteria when only outputs (or inputs) are required. Lovell and Pastor (1999) propose the pure output (or input) model without inputs (or outputs) to address this situation. They showed that an output-oriented CCR model with a single constant input and an inputoriented CCR model with a single constant output fits with the corresponding BCC models, but a CCR model with no inputs (or outputs) is meaningless. Furthermore, decisionmakers, according to their consideration or judgments, must assign weights that describe the relative importance of each MCDM criterion, while DEA enables each DMU to choose automatically the input and output weights that maximize its efficiency. Finally, MCDM techniques aim at providing a full ranking of alternatives, while the conventional DEA models do not give a ranking those DMUs deemed efficient (all of which have a score of 100%). The super DEA efficiency model, proposed by Andersen and Petersen (1993), could overcome the ranking problem. This method involves executing the standard DEA models (CRS or VRS), but under the assumption that the DMU being evaluated is excluded from the reference set so that efficient DMUs may have efficiency scores larger than or equal to 100%.

## 3. THE PROPOSED METHODOLOGY

The proposed approach is based on three stages, in which each one of them requires performing several steps (Figure 2).

### 3.1 Stage1: Fuzzy Evaluation of Perceived Quality under SERVPERF Model

Step 1: Identification of service quality criteria

Based on a literature review and for practical consideration, this study incorporates only twelve service attributes into four service dimensions of the SERVPERF model (Table 1). The two service dimensions, Responsiveness and Empathy, are aggregated to one dimension because several authors have used attributes belonging to them interchangeably.

Step 2: Fuzzification of linguistic terms

The bus users were asked to rate the importance level, and the perceived level of various service attributes using two sets of linguistic terms respectively ({Not important, slightly important, moderately important, Important, very important}; {Very dissatisfied, dissatisfied, neutral, satisfied, very satisfied}). In the fuzzy set theory, each linguistic term is transformed into a fuzzy number. The triangular numbers used in this paper, distributed between zero and ten, as illustrated in Table 2.

Step 3: Calculation of the fuzzy weighted quality perception for each bus routes with respect to every service dimension

The calculation of fuzzy weighted quality perception needs performing three sub-steps as follows:

Step 3.1: Calculation of average fuzzy importance scores and fuzzy importance weights for each service attributes:

Let lI mI uI be the TFN representing the importance level expressed by jth user (j=1,2,....,J) regarding the kth service attribute belonging to the Dth service dimension ($k = 1 , 2 , 3 , … , D k$). $( I ˜ d , k , j )$ is expressed according to the linguistic terms of the fuzzy evaluation scale represented in Table 1. Based on the arithmetic mean to aggregate group evaluation, the average importance level for each kth service attribute is given below:

$( I ˜ D , k ) = ( l D , k = ∑ j = 1 J l I D , k , j J ; m D , k = ∑ j = 1 J m I D , k , j J ; u D , k = ∑ j = 1 J u I D , k , j J )$
(7)

Then, the fuzzy importance weight of each service attribute is given as follow:

(8)

Step 3.2: Calculation of average fuzzy quality perception of each bus route with respect of every service attributes

Let $( P ˜ D , k , j ) r = ( l P D , k , j , m P D , k , j , u P D , k , j ) r$ be the TFN representing the quality perception of the bus route r expressed by the jth customer with reference to kth service attribute of the Dth service dimension. $( P ˜ D , k , j ) r$ are expressed according to the linguistic terms of the fuzzy evaluation scale represented in Table 1. Based on the arithmetic mean to aggregate group evaluation, the average perceived quality for the rth bus route is given as follows:

$( P ˜ D , k ) r = ( ( l P D , k ) r = ∑ ​ j = 1 J l P D , k , j J ; ( m P D , k ) r = ∑ ​ j = 1 J m P D , k , j J ; ( u P D , k ) r = ∑ ​ j = 1 J u P D , k , j J )$
(9)

Step 3.3: The fuzzy weighted quality perception of each bus route with respect to every service dimension

The fuzzy weighted quality perception of Dth service dimension $( g ˜ D )$ for each bus route can be obtained by aggregating the fuzzy weighted perception of its kth service attribute as reported below:

$( g ˜ D ) r = ( ∑ D = 1 D ( w ˜ D , k ) × ( P ˜ D , k ) r )$
(10)

Step 4: Calculation of the crisp quality perception scores for bus routes with respect to each service dimension

Finally, the crisp scores of the quality perception for the rth bus route as regarded to service dimension can be obtained, respectively, with reference to the confidence level (α-cut) and the index of optimism on fuzzy results μ (Lee, 1995), as follows:

$( g α D ) r = μ ⋅ ( u α g D ) r + ( 1 − µ r ) ⋅ ( l α g D ) r , ∀ r = 1 , 2 , … , R α , μ ∈ [ 0 , 1 ]$
(11)

The value of α indicates the degree of confidence in the fuzzy assessment given by the respondents with respect to service attribute importance and satisfaction ratings. A larger value is used when users are more confident in choosing a crisp value to represent their assessments. On the other hand, μ reflects the user’s attitude towards risk on fuzzy assessment results, which may be optimistic, moderate, or pessimistic. An optimistic respondent is prone to prefer higher crisp values derived from fuzzy assessments, while a pessimistic respondent to favor lower values. In our actual decision setting, μ = 0.95; 0.5 or 0.05 were considered to indicate that the user can have an optimistic, moderate, or pessimistic viewpoint, respectively, on fuzzy assessment results.

### 3.2 Stage 2: Comparative Evaluation of Perceived Quality Index (PQI)

Step 1: DEA as an MCDM Tool

The aim of this step is to provide a full ranking of bus routes given an overall measure of perceived quality. For this reason, the super efficiency DEA model, proposed by Andersen and Petersen (1993), is herein used as an MCDM technique to evaluate and fully rank bus routes (Figure 1). The effectiveness scores generated from the model are considered as the perceived quality index (PQI) for each bus route.

The super CCR DEA scores are calculated by solving the following output-oriented linear program:

(12)

Subject to:

For each of the r bus route $( r = 1 , 2 … R )$, there is a single input, $x i r ( i = 1 )$, which is a constant equal to unity (1) and set for all bus routes; and there are four o outputs , corresponding to the crisp scores of SERVPERF dimensions, i.e., tangibles, reliability, assurance, and responsiveness/empathy.

where

• - $1 / φ *$ is the effectiveness score of rj (bus route under evaluation)

• - λr represents the weight associated with the out-puts and inputs of the rj

• - When $1 φ * ≥ 100 %$ the bus route is fully effective. The larger the $1 φ *$ value is, the better the bus route is.

Step 2: Sensitivity analysis

As aforementioned, DEA outputs have been based on crisp scores. These later are calculated with reference to the confidence level and the index of optimism on fuzzy results. In this sense, a sensitivity analysis was performed to highlight the possible effects of uncertainty and ambiguity degrees of involved customers on the obtained DEA ranking results. Particularly, nine different scenarios were considered by varying the customer’s confidence level α (α = 0.05, 0.5 and 0.95) and their degree of optimism μ (μ = 0.05, 0.5 and 0.95) as well.

### 3.3 Stage 3: Benchmarking Quality Effectiveness

Step 1: Calculation of effectiveness scores

The perceived quality index, calculated in stage 2, represents only the passenger’s viewpoint. It should be matched with the level of quality as delivered by company for a more reliable analysis of quality. The proposed DEA model in this stage measures the quality effectiveness, whereby “quality effectiveness,” we mean the effectiveness with which the quality, as provided by company, is translated into the highest possible level of perceptions by costumers. The translation process is measured by an output-oriented CCR model in which the offered quality and perceived quality represent the input and output sides, respectively (Figure 4). The model provides information about how much overall perceived quality could be improved if the actual level of provided quality is respected. Also, it indicates which elements of provided service need further improvements in order to achieve a high level of overall perceived quality. The effectiveness scores are calculated using the DEA model (07).

The DEA scores are calculated by solving the fol-lowing linear program:

$Max φ + ε ( ∑ i = 1 m s i − + ∑ o = 1 s s o + ) Subject to: ∑ r = 1 R x i r λ r + s i − = x i j i = 1 , 2 , ⋯ , m ∑ r = 1 R y o r λ r + s o + = φ y o j o = 1 , 2 , ⋯ , s λ r ≥ 0 r = 1 , 2 , ⋯ , R$
(13)

Subject to:

For each of the r bus route (r = 1, 2, …, R) , there are i inputs xir (i=1, 2,…, m) , corresponding to the offered quality measures, and there are o outputs yor (o=1, 2,…, s) , corresponding to the perceived quality measures.

where

• - $φ * = 1 / φ$ is the effectiveness score of rj (bus route under evaluation)

• - $s r +$ Outputs slacks which indicate the need for further increase in corresponding outputs

• - $s i −$ Inputs slacks which signal any additional decrease of inputs

• - λj represents the weight associated with the outputs and inputs of the rj

• - ε is a non-Archimedean element smaller than any positive real number

• - rj is fully (100%) effective if and only if $φ * = 100 %$ and $s o + = s i − = 0$ for all i and o.

• - If ($φ * < 100 %$), then rj is ineffective. i.e., this route can either increase its output levels or decrease its input levels.

Step 2: Calculation of potential improvements

The DEA method defines not only the levels of input- output slacks but also the levels of input-output targets, which are the projected values on the frontier. From the equation of model (13), the levels of inputs-outputs targets are defined by the following formulas:

(14)

(15)

where,

• is the input target i for the route rj,

• yoj is the output target r for the route rj ;

• xij is the actual input i for the route rj ,

• yoj is the actual output r for the route rj ,

• $φ *$ is the effectiveness score for the route rj ;

• $s i − *$ is the optimal input slacks and $s o + *$ is the optimal output slacks.

Potential improvements indicate the percentage that a route rj needs to increase its outputs or reduce its inputs to become 100% effective. These percentages are calculated as follows:

(16)

(17)

## 4. EMPIRICAL STUDY

Some academic studies have focused on analyzing the transport sector in Fez City such as problems of public transport (e.g., Moufad and Jawab, 2017), the performance of freight transport (e.g. Moufad and Jawab, 2019) and the location of loading/unloading spaces for urban freight (e.g. Imane and Fouad, 2019). However, only a few studies are dedicated to analyzing the service quality of the bus transport service. For instance, one that explores improvement priorities (Zehmed and Jawab, 2020) and the other (Zehmed and Jawab, 2019) which lays the theoretical foundations for the present study.

In this study, we applied the three-stage approach, described above, in order to measure and benchmark the service quality of a sample of 25 bus routes in the Fez City. The bus public service is provided, under delegated management, by a single private company which covers a network of 439 Km. The vehicle fleet composed of 256 buses providing service on 51 routes connecting the principal axes of the city (Zehmed and Jawab, 2019).

### 4.1 Data

- Perceived quality data (stage 1 -2):

A sample survey was addressed to the bus users to get their perception of service quality. The survey was realized between May and June 2019, and data collection took place in the bus stops/stations of the main avenues of the city where the majority of the routes intersect. Based on a structured questionnaire, each user is asked to rate the degree of importance he/she gives to each attribute of quality. Also, he/she is requested to rate at least two routes he/she attends the most (each route independently) by specifying his/her degree of satisfaction with each service attribute. The questionnaire also gathered information about the demographic profile and travel behavior of users.

A total of 784 users were interviewed, giving 1217 responses about 44 bus routes. Excluding per urban routes and routes for which the number of responses is less than 20 responses yielded a final set of data including 25 routes with 750 users giving 1090 answers. 57% of respondents are female, while 43% are Male. The major part (73%) of interviewed passengers is aged less than 24 years, while the remaining (27%) respondents aged more than 24 years. Most (81%) of participants are students, and a considerable part (13%) is composed of the employee. The remaining part is constituted by self-employed persons (3%) and unemployed ones (3%). More than half of the sample (63%) use the bus at least once a day while 23% of interviewed persons use the bus at least once a week, and the remaining (14%) passengers use the bus at least once a month. Most (80%) of the sample people use the bus to reach places of study, and a considerable part (14%) use the bus to reach areas of work. The remaining 6% is spread among places of entertainment (2%), healthcare (1%), shopping (1%), or something else (1%).

From demographics, we notice that the majority of participants are students and are less than 24 years old. These results are reasonable because over 60% of the company customers are students (Lidia, 2014). Furthermore, according to Eboli and Mazzulla, “students are a relevant part of total public transport users because the other user categories scarcely travel by public transport” (2008, p.521). Therefore, several scientific studies about the evaluation of service quality of bus transport have been based on student samples (e.g., Eboli and Mazzulla, 2007, 2008, 2009;Cirillo et al., 2011;Güner, 2018).

- Quality effectiveness data (stage 3):

The single output of the quality effectiveness model is the perceived quality index (PQI) based on the moderate viewpoint scores, as calculated in the second stage. The inputs are objective indicators that reflect the level of the offered quality. However, the bus operator at Fez City has not yet made any evaluation of service quality at the route level. There are no statistics about the regularity of runs or punctuality of buses. There are also no statistics about road accidents, crimes on-bus, or crimes at bus stops. Instead, the bus operator provides us only operational data that reflect the main characteristics of service. Hence, in this study, average headway (Minutes), stop spacing (Km), and average travel time (Minutes) are used as measures of offered quality. Average headway (minutes) is the average time interval between transit vehicles. It’s a reciprocal measure of service frequency (transit vehicles per hour) and indicates the level of availability of service (Kittelson & Associates, Inc. et al., 2013). Stop spacing (Km) is the average distance between stops, and it measures the level of accessibility. It’s obtained by dividing the length of the bus route by the total number of stops (Tran et al., 2017). Finally, the average travel time (Minutes) is the average travel time taken by a one-way bus ride from departure to an arrival station. It’s a measure of comfort (Kittelson & Associates, Inc. et al., 2013). All these measures reflect service attributes (accessibility, availability, and comfort) that are found to be most influential in overall perceived quality in several studies (e.g., Mouwen, 2015;Morton et al., 2016;Mahmoud and Hine, 2016) and are among the most important service attributes for users in fez city (Zehmed and Jawab, 2020). Furthermore, these measures are the most used as input variables in the bus route efficiency evaluation with the DEA method (Karim and Fouad, 2018). Table 3 provides a statistical summary of inputs data.

### 4.2 Results and Discussion

#### 4.2.1 Fuzzy Weighted Scores of SERVPERF Dimensions

The fuzzy average importance scores and the fuzzy importance weights of each service attribute are calculated using formula (07) and (08), respectively. Next, formula (09) is used to calculate the fuzzy quality perception scores of each bus route for every service attribute. For reasons of space, results of formulas (07), (08), and (09) are presented in the appendix 1 & 2. Finally, formula (10) is applied to calculate the fuzzy weighted quality perceptions of each bus route as regards to each service dimension (Table 4). The results show unsatisfactory levels of perceived quality in almost bus routes except route #9, which obtained a medium level in terms of “Assurance” and “Responsiveness/Empathy.”

#### 4.2.2 Crisp Scores of SERVPERF Dimensions

Eq.11 was applied to calculate the crisp quality perception scores of each bus route as regards to every service dimension (Table 5). In all these equations, both the confidence level and the index of optimism were assumed equal to 0.5 (moderate viewpoint). The results, which are presented in table 5, showed that ranking varies in a different sense, which will not allow us to make a valid comparison. For this reason, the DEA method was then employed as an MCDM technique to get a composite indicator of the perceived quality for each bus route.

#### 4.2.3 PQI Results and Sensitivity Analysis

Applying the model (12), the final scores and ranking of the 25 bus routes are given in Table 6. From results, we notice that only bus routes #9 and #54 are fully effective in terms of perceived quality. To test the stability of results, we performed a sensitivity analysis as described in stage 2-step 3 of the methodology section. The results, as reported in Table 4, showed that overall perceived quality scores were characterized by constant values with respect to the confidence level α as well as there were no ranking-reversal situations regards to transitions from pessimistic to optimistic viewpoints.

#### 4.2.4 Quality Effectiveness Scores

Reminding that the output of quality effectiveness DEA model is based on PQI (moderate viewpoint), the model (13) is applied and results are given in table 8. Second column shows the effectiveness scores and the third column shows the reference and their weight for each ineffective bus route. The weights represent the contribution of each peer on its reference. For example, #42 has an effectiveness score of 73.10%, and it is considered ineffective. #15 and #52 are assigned as references to #42 in order to carry it on the isoquant. Since the weight of #52 is higher than #15 ((1.247073) and (0.001561) respectively), #52 is more influential on #42 in the benchmarking process than #52.

The results show that of the 25 bus routes, only five routes (#12, #15, #29, #52 and #9) are fully effective while the remaining routes can be considered as ineffective with a degree that ranges from 2.19% to 49.18%. Furthermore, the quality effectiveness score ranges from 50.82% to 100%, and the average score is 80.57%. Finally, we note that route #9 and #52 are the major references for most of the ineffective routes.

#### 4.2.5 Percentage of Potential Improvements

The 20 ineffective bus routes can improve their effectiveness by adjusting the level of perceived quality and level of offered quality while comparing with their references. Using formulas (16) and (17), table 9 provides the percentage of potential improvements in each bus route. The results show that the overall perceived quality will increase in 9 routes (#11, #16, #2, #20, #23b, #30, #31, #41, #51) if the actual level of service quality is strictly respected. The most significant increase is noticed in bus routes #16, #2, and #41, while the minor changes are observed in routes #50, #20, and #11. To achieve a high level of perceived quality, more adjustments to the level of offered service quality are needed in the other routes. For instance, the results suggest changes in five bus routes (#13, #14, #25, #51 and #53), in terms of stop spacing, where the most critical reduction is noticed in bus route #51. Concerning travel time, decreases are suggested in seven bus routes (#17, #23, #25, #4, #42, #5, and #51), where the most important reduction is observed in route #17. Finally, only three routes (#17, #4, and #5) need also reductions concerning headway time. The most significant decrease is noticed in bus route #4.

#### 4.2.6 Practical Implications and Recommendations

The results of the study have many practical implications: The majority of bus routes showed unsatisfactory levels of perceived quality except routes #9 and #54 that showed medium levels. This result should be taken seriously by the bus operator managers because it is not sensitive to the subjectivity of the respondents (e.g., from pessimistic to optimistic). Ideally, the service quality should be improved over different dimensions in almost bus routes, considering routes #9 and #54 as benchmarks. However, the implementation of all these improvements generates more costs and needs more investments. For this reason, an interaction between the provided level of quality by the operator and the overall perceived quality has been established in the third stage. The results provide important insights for bus operator managers about how much overall perceived quality could be improved if the actual level of delivered quality is respected. In addition, it indicates which elements of provided service need further improvements in order to achieve a high level of overall perceived quality by customers.

The provided quality is reflected in this paper by three aspects, namely, accessibility (Stop spacing), availability (Headway), and comfort (Travel Time). The improvement of geographic accessibility could be made by revising the location of stops, route length, and itinerary of the route. The temporal availability could be improved by increasing the frequency of runs and assigned additional buses to concerned routes. Travel time is dependent on the operating environment as the bus service in Fez city is operated in a mixed traffic route. The improvement of travel time could be made by giving priority to buses at signalized intersections or implement exclusive bus lanes when route conditions permit. Additionally, the implementation of a real-time passenger information system will ensure better-provided service quality. Finally, it is worth noting that the successful implementation of all the above improvements requires a commitment and fruitful collaboration between the delegating authority (Fez City) and the private bus operator.

## 5. CONCLUSION

Due to its essential role in urban public transport, service quality has been evaluated in scientific research using different methodologies and under different points of view. In this paper, a three-stage approach based on the fuzzy SERFPERF and Data Envelopment Analysis (DEA) method have been proposed to evaluate and benchmark the bus transport service at the route level. In the first stage, a fuzzy evaluation of perceived quality under the SERVPERF model has been done. Namely, triangular fuzzy numbers are considered to calculate the average perceived quality of each bus route with respect to SERVPERF dimensions. In the second stage, the super- DEA method has been used as an MCDM tool to aggregating the dimension’s scores into a single measure of perceived quality. In addition, a sensitivity analysis was performed to highlight the possible effects of uncertainty and ambiguity degrees of involved customers on the obtained super-DEA results. In the third stage, a DEA quality effectiveness model is suggested where the company and customers’ points of view are, respectively, taking into account. Namely, the model identifies bus routes where high levels of perceived quality could not be achieved given the level of offered quality by the bus company, and provide potential improvements for these routes. Finally, an empirical study was conducted concerning a sample of 25 bus routes in a major Moroccan city named Fez.

The most important contributions of the study are twofold. First, this study has demonstrated a new application of DEA as an MCDM tool in the context of public transport services at the route level. Namely, DEA has been used to construct a comprehensive measure of perceived quality from fuzzy SERVPERF dimensions scores. Second, it demonstrated the feasibility of DEA in benchmarking service quality while establishing an interaction between the company’s and passengers’ points of view.

The present paper also has some limitations that future research should address. First, the quality effectiveness model established interaction only between an aggregate score of perceived quality with objective measures that are limited in terms of service aspects that they represent. However, it will be more beneficial for quality management if running separates DEA models. Each model establishes the interaction between one perceived SERVPERF dimension with its corresponding objective measures. Second, the passenger’s point of view about quality is limited to their perceptions. Nevertheless, the consideration of the quantitative factor quality such as passenger-miles will give more significant results.

## Figure

Triangular fuzzy number Ã

Diagram of the proposed methodology.

DEA as an MCDM tool.

Conceptual DEA Model for quality effectiveness of the bus route.

## Table

Service dimensions and attributes of the perceived quality for bus route transport

Default values of linguistic terms represented as triangular fuzzy numbers

The descriptive inputs data

Fuzzy weighted scores of SERVPERF dimensions

Fuzzy Importance Weights of Each Service Attributes

Fuzzy Average Perceived Quality for Each Bus Route as Regards to Service Attributes (formula: (09))

Crisp scores and ranking of bus routes

Scores and ranking of bus routes (α=0.5 & u=0.5)

Bus routes scores and ranking in different scenarios

Effectiveness scores and references

Percentage of potential improvements

## REFERENCES

1. Adler, N. and Berechman, J. (2001), Measuring airport quality from the airlines’ viewpoint: An application of data envelopment analysis, Transport Policy, 8(3), 171-181.
2. Adolphson, D. L. , Cornia, G. C. , and Walters, L. (1990), A unified framework for classifying DEA models, In: H. Bradley (ed.), Operational Research’90, Pergamon Press, New York.
3. Andersen, P. and Petersen, N. C. (1993), A procedure for ranking efficient units in data envelopment analysis, Management Science, 39(10), 1261-1264.
4. Awasthi, A. , Chauhan, S. S. , Omrani, H. , and Panahi, A. (2011), A hybrid approach based on SERVQUAL and fuzzy TOPSIS for evaluating transportation service quality, Computers & Industrial Engineering, 61(3), 637-646.
5. Aydin, N. (2017), A fuzzy-based multi-dimensional and multi-period service quality evaluation outline for rail transit systems, Transport Policy, 55, 87-98.
6. Aydin, N. , Celik, E. , and Gumus, A. T. (2015), A hierarchical customer satisfaction framework for evaluating rail transit systems of Istanbul, Transportation Research Part A: Policy and Practice, 77, 61-81.
7. Banker, R. D. , Charnes, A. , and Cooper, W. W. (1984), Some models for estimating technical and scale inefficiencies in data envelopment analysis, Management Science, 30(9), 1078-1092.
8. Barabino, B. , Deiana, E. , and Tilocca, P. (2012), Measuring service quality in urban bus transport: A modified SERVQUAL approach, International Journal of Quality and Service Sciences, 4(3), 238-252.
9. Barabino, B. and Di Francesco, M. (2016), Characterizing, measuring, and managing transit service quality: Transqual, Journal of Advanced Transportation, 50(5), 818-840.
10. Bayraktar, E. , Tatoglu, E. , Turkyilmaz, A. , Delen, D. , and Zaim, S. (2012), Measuring the efficiency of customer satisfaction and loyalty for mobile phone brands with DEA, Expert Systems with Applications, 39(1), 99-106.
11. Belton, V. and Vickers, S. P. (1993), Demystifying DEA: A visual interactive approach based on multiple criteria analysis, Journal of the Operational Research Society, 44(9), 883-896.
12. Bilişik, Ö. N. , Erdoğan, M. , Kaya, İ. , and Baraçlı, H. (2013), A hybrid fuzzy methodology to evaluate customer satisfaction in a public transportation system for Istanbul, Total Quality Management & Business Excellence, 24(9-10), 1141-1159.
13. Celik, E. , Aydin, N. , and Gumus, A. T. (2014), A multiattribute customer satisfaction evaluation approach for rail transit network: A real case study for Istanbul, Turkey, Transport Policy, 36, 283-293.
14. Charles, V. and Kumar, M. (2014), Satisficing data envelopment analysis: An application to SERVQUAL efficiency, Measurement, 51, 71-80.
15. Charnes, A. , Cooper, W. W. , and Rhodes, E. (1978), Measuring the efficiency of decision making units, European Journal of Operational Research, 2(6), 429-444.
16. Chica-Olmo, J. , Gachs-Sánchez, H. , and Lizarraga, C. (2018), Route effect on the perception of public transport services quality, Transport Policy, 67, 40-48.
17. Cirillo, C. , Eboli, L. , and Mazzulla, G. (2011), On the asymmetric user perception of transit service quality, International Journal of Sustainable Transportation, 5(4), 216-232.
18. Cronin, J. J. and Taylor, S. A. (1992), Measuring service quality: A reexamination and extension, Journal of Marketing, 56(3), 55-68.
19. de Aquino, J. T. , de Melo, F. J. C. , Jerônimo, T. de B. , and de Medeiros, D. D. (2018), Evaluation of quality in public transport services: The use of quality dimensions as an input for fuzzy TOPSIS, International Journal of Fuzzy Systems, 21, 176-193
20. de Oña, R. , Eboli, L. , and Mazzulla, G. (2014), Monitoring changes in transit service quality over time, Procedia-Social and Behavioral Sciences, 111, 974-983.
21. Doyle, J. and Green, R. (1993), Data envelopment analysis and multiple criteria decision making, Omega, 21(6), 713-715.
22. Eboli, L. and Mazzulla, G. (2007), Service quality attributes affecting customer satisfaction for bus transit, Journal of Public Transportation, 10(3), 21-34.
23. Eboli, L. and Mazzulla, G. (2008), A stated preference experiment for measuring service quality in public transport, Transportation Planning and Technology, 31(5), 509-523.
24. Eboli, L. and Mazzulla, G. (2009), A new customer satisfaction index for evaluating transit service quality, Journal of Public Transportation, 12(3), 21-37.
25. Eboli, L. and Mazzulla, G. (2011), A methodology for evaluating transit service quality based on subjective and objective measures from the passenger’s point of view, Transport Policy, 18(1), 172-181.
26. Golany, B. (1988), An interactive MOLP procedure for the extension of DEA to effectiveness analysis, The Journal of the Operational Research Society, 39(8), 725-734.
27. Güner, S. (2018), Measuring the quality of public transportation systems and ranking the bus transit routes using multi-criteria decision making techniques, Case Studies on Transport Policy, 6(2), 214-224.
28. Hassan, M. N. , Hawas, Y. E. , and Ahmed, K. (2013), A multi-dimensional framework for evaluating the transit service performance, Transportation Research Part A: Policy and Practice, 50, 47-61.
30. Kanuganti, S. , Subramanian, U. , Shrikant, S. , Singh, A. P. , and Sarkar, A. K. (2013), Quantification of level-of-service index for bus routes in developing countries: A case study in India, Journal of the Eastern Asia Society for Transportation Studies, 10, 1347-1366.
31. Karim, Z. and Fouad, J. (2018), Measuring urban public transport performance on route level: A literature review, MATEC Web of Conferences, 200, 00021.
32. Khalili-Damghani, K. , Taghavi-Fard, M. , and Karbaschi, K. (2015), A hybrid approach based on multi-criteria satisfaction analysis (MUSA) and a network data envelopment analysis (NDEA) to evaluate efficiency of customer services in Bank Branches, Industrial Engineering and Management Systems, 14(4), 347-371.
33. Kittelson & Associates, Inc., Brinckerhoff, P., KFH Group, Inc., Texas A&M Transportation Institute, Arup, Transit Cooperative Research Program, Transportation Research Board, & National Academies of Sciences, Engineering, and Medicine (2013), Transit capacity and quality of service manual (3rd ed.), Transportation Research Board.
34. Lee, A. R. (1995), Application of modified fuzzy ahp method to analyze bolting sequence of structural joints, Ph.D. thesis, Lehigh University, Bethlehem, PA, USA.
35. Lee, H. and Kim, C. (2014), Benchmarking of service quality with data envelopment analysis, Expert Systems with Applications, 41(8), 3761-3768.
36. Lewis, H. F. and Mazvancheryl, S. K. (2011), A model for efficiency analysis of the customer satisfaction process, Innovative Marketing, 7(1), 33-45.
37. Lidia, S. (2014), État des lieux synthétique du transport public dans les villes membres du REMA-TP (p. 52). CoMun-coopération municipal: Gouvernance locale et participative au Maghreb, Available from: http://co-mun.net/maroc/les-reseaux-thematiques/rema-tp-transport-public/etat-des-lieux-et-benchmarking.
38. Lin, J. , Wang, P. , and Barnum, D. T. (2008), A quality control framework for bus schedule reliability, Trans­portation Research Part E: Logistics and Transportation Review, 44(6), 1086-1098.
39. Lovell, C. A. K. and Pastor, J. T. (1999), Radial DEA models without inputs or without outputs, European Journal of Operational Research, 118(1), 46-51.
40. Mahmoud, M. and Hine, J. (2016), Measuring the influence of bus service quality on the perception of users, Transportation Planning and Technology, 39(3), 284-299.
41. Mainardes, E. , Alves, H. , and Raposo, M. (2014), Using expectations and satisfaction to measure the frontiers of efficiency in public universities, Tertiary Edu­cation and Management, 20(4), 339-353.
42. Mohanty, P. K. and Senthil Kumar, N. (2017), Measuring farmer’s satisfaction and brand loyalty toward Indian fertilizer brands using DEA, Journal of Brand Ma­nagement, 24(5), 467-488.
43. Morton, C. , Caulfield, B. , and Anable, J. (2016), Cus­tomer perceptions of quality of service in public transport: Evidence for bus transit in Scotland, Case Studies on Transport Policy, 4(3), 199-207.
44. Moufad, I. and Jawab, F. (2017), Multi-criteria analysis of urban public transport problems: The city of Fes as a case, International Journal of Scientific & Tech­nology Research, 8(1), 675-682.
45. Moufad, I. and Jawab, F. (2019), A study framework for assessing the performance of the urban freight transport based on PLS approach, Archives of Transport, 49(1), 69-85.
46. Mouwen, A. (2015), Drivers of customer satisfaction with public transport services, Transportation Research Part A: Policy and Practice, 78, 1-20.
47. Najafi, S. , Saati, S. , and Tavana, M. (2015), Data envelopment analysis in service quality evaluation: An empirical study, Journal of Industrial Engineering International, 11(3), 319-330.
48. Nikel, C. , Eldeeb, G. , and Mohamed, M. (2020), Perceived quality of bus transit services: A route-level analysis, Transportation Research Record: Journal of the Transportation Research Board, 2674(2), 79-92.
49. Parasuraman, A. , Zeithaml, V. A. , and Berry, L. L. (1985), A conceptual model of service quality and its implications for future research, Journal of Marketing, 49(4), 41-50.
50. Parasuraman, A. , Zeithaml, V. A. , and Leonard, L. B. (1988), SERVQUAL: A multiple-item scale for measuring consumer perceptions of service quality, Journal of Retailing, 64(1), 12-40.
51. Podinovski, V. V. and Thanassoulis, E. (2007), Improving discrimination in data envelopment analysis: Some practical suggestions, Journal of Productivity Analysis, 28(1-2), 117-126.
52. Sam, E. F. , Hamidu, O. , and Daniels, S. (2018), SERVQUAL analysis of public bus transport services in Kumasi metropolis, Ghana: Core user perspectives, Case Studies on Transport Policy, 6(1), 25-31.
53. Soteriou, A. and Zenios, S. A. (1999), Operations, quality, and profitability in the provision of banking services, Management Science, 45(9), 1221-1238.
54. Stewart, T. J. (1996), Relationships between data envelopment analysis and multicriteria decision analysis, Journal of the Operational Research Society, 47(5), 654-665.
55. Sun, D. (Jian) , Chen, S. , Zhang, C. , and Shen, S. (2016), A bus route evaluation model based on GIS and super-efficient data envelopment analysis, Transportation Planning and Technology, 39(4), 407-423.
56. Tran, K. D. , Bhaskar, A. , Bunker, J. M. , and Lee, B. L. (2017), Data Envelopment Analysis (DEA) based transit routes performance evaluation, Proceedings of the Transportation Research Board (TRB) 96th Annual Meeting. Transportation Research Board (TRB), United States of America, Washington D.C., 1-23.
57. Wojcik, V. , Dyckhoff, H. , and Clermont, M. (2019), Is data envelopment analysis a suitable tool for performance measurement and benchmarking in non-production contexts? Business Research, 12(2), 559-595.
58. Yeh, C. H. , Deng, H. , and Chang, Y. H. (2000), Fuzzy multicriteria analysis for performance evaluation of bus companies, European Journal of Operational Research, 126(3), 459-473.
59. Yu, M. C. , Keng, I. , and Chen, H. X. (2015), Measuring service quality via a fuzzy analytical approach, International Journal of Fuzzy Systems, 17(2), 292-302.
60. Zadeh, L. A. (1965), Fuzzy sets, Information and Control, 8(3), 338-353.
61. Zehmed, K. and Jawab, F. (2020), Exploring improvement priorities for urban bus service: A network and routes perspectives, International Journal of Scientific & Technology Research, 9(01), 2119-2124.
62. Zehmed, K. and Jawab, F. (2019), Analyzing the quality of public bus transport service in Fez City (Morocco), Proceeding of the 12th International Colloquium of Logistics and Supply Chain Management LOGISTIQUA 2019, Montreuil - Paris, France, 1-6.
 Do not open for a day Close