 Research
 Open Access
Multicriteria ranking of voice transmission carriers of a telecommunication company using PROMETHEE
 Beatrice Boatemaa^{1},
 Justice K. Appati^{2}Email authorView ORCID ID profile and
 Kwaku F. Darkwah^{1}
 Received: 12 August 2018
 Accepted: 9 November 2018
 Published: 22 November 2018
Abstract
This paper ranks the voice transmission carriers as used by mobile telecommunication companies. The versatility of the PROMETHEE method made it a useful tool for this ranking process. However, in our approach, the logistic preference function which is a recently proposed preference function was adopted in the ranking procedure as opposed to the Gaussian preference function. The results obtained by the logistic preference function yield a similar effect as that of the Gaussian preference function with both reaching an optimal ranking at the complete ranking stage.
Keywords
 Complete ranking
 Logistic preference function
 Mobile telecommunication
 Multicriteria decision making
 PROMETHEE method
Introduction
Telecommunication carrier is by definition companies that are authorized by a regulatory agency to operate a telecommunications system. In the work of ChingTer et al. (2011), the problem of routing management still stands to confuse many telecommunicationrelated companies among which are: equipment manufacturers, platform vendors, service operators, and billing system. Through their study, they observed that routing control is dependent not only on equipment but also on the operational flow of the company. Unlike Internet Service Provider (ISP) where its providers often control the entire network topology; the same is not of telecommunication operators. A full range of service is only possible to the reach of its customers by contracting partners such as service carriers. In the case of customer calls or user calls, these operators provide service according to an internal and predeployed routing logic to its service providers or carriers depending upon the destination or the service traffic to available routes. That is, a clear guide is instituted for the engineering staff on path deploying for the next period (or billing cycle) to come while achieving multiple predefined goals automatically. In this instance, decisionmaking which is the science of recognizing and determining choices based on the values and partialities of the decision maker becomes necessary (Diaz et al. 2015; Caron and Daniels 2013; Anderson and Mansingh 2014). Making a choice implies that there are alternatives to be considered, and in such a case we do not only distinguish as many of these choices but also take the one that best fits with the objectives, values, goals, and desires. This process starts with the classification of the decision makers and stakeholders in the resolution and reducing the possible misunderstanding about problem definition, requirements, criteria, and goals (Majumder and Ghosh 2013; AlTarawneh 2012). According to Ignaccolo et al. (2017), Moffett and Sarkar (2006), to make a decision, one needs to understand the problem, the urgency, the scope of the decision, the criteria for the choice, their subcriteria, stakeholders, and groups affected and the alternative steps to take. In the case of resource allocation, one requires the needed priorities to determine the best alternative to allocate the appropriate share of the resources. They also recommended that the decisionmaking process must distinguish causes, limiting assumptions, policy and organizational barriers and interfaces, and any stakeholder concerns. The aim is to reveal the issue in a fair and precise problem statement that represents both the first circumstances and the desired circumstances. The problem statement must nevertheless be a brief and unambiguous written material agreed upon by all decisionmakers and stakeholders.
Related works
In Triantaphyllou and Mann (1995), Nemhauser et al. (1989), it was explained that decisionmaking suggests many criteria and subcriteria employed to order the options of choice. Not solely does one need to generate preferences for the options about the requirements for which they must be estimated, but also regarding essential goals. The works of Steuer (1986), Steuer and Na (2003), Hirschberger et al. (2013) also stipulated that the case where we have a finite number of criteria but the number of feasible alternatives are infinite belongs to the field of multiple criteria optimization. Contrary to that, decisionmaking problems where the number of the criteria and alternatives is finite with an explicit definition of alternatives are called multiattribute decisionmaking problems. Again, Kuhn and Tucker (1951) considered problems with multiple objectives while formulating optimality conditions for nonlinear programming. The Multiple Objective Programming (MOP) problem seeks to optimize a set of incommensurable and conflicting objectives under some constraints simultaneously and this was brought to bear in Charnes et al. (1955), Charnes and Cooper (1961), Contini and Zionts (1968). In Saaty (1977, 1986, 1990, 2003), Analytic Hierarchy Process (AHP) which is a multicriteria decisionmaking approach was introduced with tools that can be used to solve complex decision problems. These have captivated the curiosity of many researchers due to their critical mathematical properties and the requirement that the input data are easy to get. According to Merkin (1979) AHP is based on the welldefined mathematical structure of uniform matrices with correlated eigenvector’s to generate real or approximate weights. The method examines the models and alternatives concerning a criterion in a consistent and pairwise mode. The AHP uses a necessary scale of absolute numbers that have been proven and validated experimentally to capture individual preferences concerning quantitative and qualitative attributes (Saaty 1980, 1990). The study of Kumar et al. (2009) explores the analytical hierarchy process as applied to vendor selection in the following categories: small, medium and largescale industries. Their work investigates the problem associated with the use of AHP as a vendor selection tool under different criteria (Singh et al. 2003, 2005). For large companies, a formal approach is adopted in selecting suppliers as opposed to small and mediumsized industries (Pearson and Ellram 1995). The AHP method is also adopted by Tam and Tummala (2001) in vendor selection of a telecommunications system. In their article, they formulated an AHPbased model to examine the feasibility of selecting a vendor for a telecommunications system as a case study. The proposed model shows an improvement in the group decisionmaking in vendor selection that matches with customer specifications. This model gives a systematic process with the reduced time taken in vendor selection. These articles (Figueira et al. 2009, 2013) also developed ELECTRE which is a family of multicriteria decision analysis methods with the concept of constructing a directed network of preferences. With this approach, a set of outranking decisions which is considered as best is constructed. In the work of Brans and Vincke (1985), PROMETHEE I also know as partial ranking, and PROMETHEE II that is complete ranking was developed. Some years later, PROMETHEE III which is a ranking strategy based on the intervals, and PROMETHEE IV the continuous case was proposed by Brans and Mareschal (1994), Brans et al. (1986). The same authors proposed the interactive visual module GAIA, a graphical representation to support their methodology (Deshmukh 2013), while the work of Brans and Mareschal (1992) further suggest PROMETHEE V as an extension with the inclusion of the segmentation constraints. With the representation of the human brain, the work of Brans and Mareschal (1995) proposed PROMETHEE VI. The decision table which serves as the starting point of the PROMETHEE method was proposed by Brans and Vincke (1985). In the decision table is a score defined of which for simplicity sake do not require normalization or transformation to a dimensionless scale. It is, therefore, assumed that a higher score value implies better performance, while the weights of the criteria are determined by an appropriate method other than the PROMETHEE method. In the work of Anagnostopolous et al. (2005), the PROMETHEE method with the AHP was adopted in the study of water resource planning. Their scope of the study was the management of the Nestos river in relation to the operation of two dams constructed. The management of the water supply is required to create a balance in the needs for irrigation, the public electrical corporation for hydropower generation, as well as environmental requirements in the presence of valuable natural ecosystems in the area. The study of Athawale and Chakraborty (2010) adopted the PROMETHEE II method in facility location selection. This method becomes relevant to decision makers in the citation of a new organization or the expansion of existing facility. The association of cost to the acquisition of land and the construction of facility makes facility location problem a longterm investment decision. In the UK police force, the analysis of improvement of rank performance was carried out in the work of Barton and Beynon (2006a, b) using the PROMETHEE method. Their study took twofolds, where the first exploits the PROMETHEEbased uncertainty analysis in rank improvement while the second phase makes use of the result from the first phase to aid in the performance strategies. Finally, the survey of Velasquez and Hester (2013) indicated the strengths and the weaknesses of the various multicriteria methods including AHP, PROMETHEE, MAUT, and ELECTRE. In this paper, we use the logistic preference function (Amponsah et al. 2012) in the PROMETHEE method to rank voice transmission carriers of a mobile telecommunication company.
Methods
The problem of Decision Optimization is stated as follows: Let the set \(A = \{A_j\}, j = 1, \ldots , m,\) be a finite set of alternatives. Let the set \(C = \{C_i\}, i = 1, \ldots , n\) be a weighted finite set of criteria with \(w_i\) being the weights Let \(X_{ij}\) be realvalued performance index of the alternative j on criterion i, determine alternative \(A_j\) with index j as the best choice alternative satisfying the criteria The Preference Ranking Organization Method for Enrichment Evaluation (PROMETHEE) I and II were developed by Brans and Vincke (1985) to provide a ranking from best to worst of a finite set of alternatives. Ranking of alternatives based on PROMETHEE methodology requires the following steps:
 Step 1::

Input data of decision table: the data show in quantitative terms the performance of each alternative \(A_j\) on each criterion, \(C_i\) for all the criteria C. Cost and profit criteria are labeled as minimizing and maximizing criteria, respectively. The performance is denoted by the \(X_{ij}\) in the decision table.
 Step 2::

Calculate deviations of various criteria values: for all maximizing criteria we use the equationWhile for minimizing criteria we use the equation$$\begin{aligned} d_i(A_k,A_l) = X_{ik}  X_{il} \end{aligned}$$(1)$$\begin{aligned} d_i(A_k,A_l) = (X_{ik}  X_{il}) \end{aligned}$$(2)
 Step 3::

Select a generalized preference function: Podvezko and Podviezko (2010) put forward eight generalized preference functions. These generalized preference functions include
 1.
The ushape preference function,
 2.
The level preference function,
 3.
The multistage preference function,
 4.
The vshape preference function,
 5.
The vshape with indifference preference function,
 6.
The cshape preference function and
 7.
The Gaussian preference function.
where \(\sigma \) is the standard deviation. The logistic preference function is used for the computations in this paper.$$\begin{aligned} P(d) = \left\{ \begin{array}{ll} 0 &{} d \le 0\\ \frac{1  e^{\frac{d^2}{\sigma ^2}}}{1 + e^{\frac{d^2}{\sigma ^2}}} &{} d > 0\\ \end{array} \right. \end{aligned}$$(3)  1.
 Step 4::

Calculate the preference (criterion) value denoted by \(P_i(d_i(A_k,A_i)).\) Using the logistic preference function, we have:The calculated value \(P_i(d_i(A_k,A_l))\) measures the intensity of the decision maker’s preference for the alternative \(A_k\) over \(A_i\) on the same criterion \(C_i\) for \(i = 1, \ldots , n\).$$\begin{aligned} P_i(d_i(A_k,A_l)) = \left\{ \begin{array}{ll} 0 &{} d \le 0\\ \frac{1  e^{\frac{d^2}{\sigma ^2}}}{1 + e^{\frac{d^2}{\sigma ^2}}} &{} d > 0\\ \end{array} \right. \end{aligned}$$(4)
 Step 5::

Calculate the preference index: we calculate the preference index for each alternative \(A_k\) over another alternative \(A_i\) for all criteria \(C_i\) by using the relationwhere \(w_i\) is the weight of criterion i.$$\begin{aligned} \pi (A_k,A_i) = \sum _{i=1}^n w_ip_i(A_k,A_i) \quad \hbox { for } k,l = 1, \ldots , m \end{aligned}$$(5)
 Step 6::

Rank the alternatives: ranking a finite set of alternatives in PROMETHEE may involve two main stages namely,
 1Partial ranking: this is known as PROMETHEE I and it establishes the outranking relation existing between various alternatives via the leaving (positive) and the entering (negative) flows \((\phi ^+(A_k))\), \((\phi ^(A_k))\) respectively.
 (a)Positive outranking flow for \(A_j\) is the preference of the alternative \(A_j\) over all other alternatives \(A_k\) and \(A_k \ne A_j\). The flow is denoted by$$\begin{aligned} \phi ^+(A_j) = \frac{1}{m1}\sum _{k=1}^m \pi (A_j,A_k), \quad j = 1, \ldots , m \end{aligned}$$(6)
 (b)Negative outranking flow for \(A_j\) is the preference of all other alternatives \(A_k\) over \(A_j\in A\) and \(A_k \ne A_j\)$$\begin{aligned} \phi ^(A_j) = \frac{1}{m1}\sum _{k=1}^m \pi (A_k,A_j), \quad j = 1, \ldots , m \end{aligned}$$(7)
 (a)
 2Complete ranking (PROMETHEE II): at this stage, it is the PROMETHEE II (preorder complete ranking) which completes the whole ranking process, establishing a relation that links all alternatives, be they comparable or incomparable and placing them in their right perspective, in a hierarchy from best to worst. Therefore, if after partial ranking, some alternatives are found to be incomparable, then we apply PROMETHEE II method to finish the ranking process for an optimal decision to be made. In fact, it makes use of only the p and I (preference and indifference respectively). This approach makes use of what is called the net outranking flow symbolically represented by \(\phi (A_k)\) for the alternative \(A_k\) such thatthe alternative \(A_l\) in terms of the net outranking flow becomes:$$\begin{aligned} \phi (A_k) = \phi ^+(A_k)  \phi ^(A_k) \end{aligned}$$(8)So, the higher the net flow the better the alternative.$$\begin{aligned} \phi (A_l) = \phi ^+(A_l)  \phi ^(A_l)\quad \forall A_l \in A \end{aligned}$$(9)Table 2 shows the two (2) existing relations between alternatives in complete ranking

(i) The alternative \(A_k\) is preferable to \(A_l\) if and only if \(\phi (A_k) > \phi (A_l)\)

(ii) The alternative \(A_k\) is indifferent to \(A_l\) if and only if \(\phi (A_k) = \phi (A_l)\)

Table 1Relations between alternatives in PROMETHEE partial preorder ranking
Preference relation
Cases
Graphical representation
\(A_kPA_l\)
\(\phi ^+(A_k) > \phi ^+(A_l)\) and \(\phi ^(A_k) < \phi ^(A_l)\)
\(A_k \rightarrow A_l\)
\(\phi ^+(A_k) > \phi ^+(A_l)\) and \(\phi ^(A_k) = \phi ^(A_l)\)
\(\phi ^+(A_k) = \phi ^+(A_l)\) and \(\phi ^(A_k) < \phi ^(A_l)\)
\(A_kIA_l\)
\(\phi ^+(A_k) = \phi ^+(A_l)\) and \(\phi ^(A_k) =\phi ^(A_l)\)
–
\(A_kRA_l\)
\(\phi ^+(A_k) > \phi ^+(A_l)\) and \(\phi ^(A_k) > \phi ^(A_l)\)
–
\(\phi ^+(A_l) > \phi ^+(A_k)\) and \(\phi ^(A_l) > \phi ^(A_k)\)
Table 2The two existing relations between alternatives in complete ranking
Preference relation
Cases
Graphical representation
\(A_kPA_l\)
\(\phi (A_k) > \phi (A_l)\)
\(A_k\rightarrow A_l\)
\(A_kIA_l\)
\(\phi (A_k) = \phi (A_l)\)
–
 1
Data
 1Route availability (\(C_1\)): The logical (trunk line) connection between two switch or exchange nodes define route of calls. Route availability is defined as:where NTO is the number of trunk outage, TNT is the total number of trunks. Route availability should not be less than 100%, hence it is a maximizing criterion$$\begin{aligned} \hbox {Route availability}[\%] = \frac{\hbox {NTO}}{\hbox {TNT}}\times 100 \end{aligned}$$(10)
 2Busy hour traffic (\(C_2\)): In a day, the 60 min interval in which the traffic is highest is called busy hour (BH). If the average number of calls to and from a terminal during a period of T second in traffic hour is ‘n’ and the average holding time before personal calls end is ‘h’ seconds, the average occupancy of the terminal is given by:The busy hour traffic should be high and, therefore, it is a maximizing criterion.$$\begin{aligned} A = \frac{nh}{T} \end{aligned}$$(11)
 3
Capacity (\(C_3\)): The capacity of a given carrier is measured in terms of the subscribers or the traffic load that it can handle. The Erlang B formula is the most commonly used figure in any telecommunication capacity calculation, (Chromy 2011). The capacity of the carrier should be high and thus it is a maximizing criterion.
 4Utilization (\(C_4\)): The utilization of the trunk is calculated as a ratio of the total traffic to the capacity expressed as a percentage:The utilization should not exceed 80%, it is, therefore, a minimizing criterion$$\begin{aligned} \hbox {Utilization}[\%] = \frac{\hbox {Total traffic}}{\hbox {Capacity}}\times 100 \end{aligned}$$(12)
 5Congestion (\(C_5\)): It is the condition in a switching center when a caller cannot obtain a connection to the wanted end user immediately. It is expressed as:where NCFC is the no. of connected failed calls, TNCA is the total No. of call attempts. Congestion should be less than or equal to 1%, it is a minimizing criterion.$$\begin{aligned} \hbox {Congestion}[\%] = \frac{\hbox {NCFC}}{\hbox {TNCA}}\times 100, \end{aligned}$$(13)
 6
Answer to seizure ratio (\(C_6\)): The answer/seizure ratio (ASR) is a measurement of network quality and call success rate in telecommunications. It is the percentage of answered telephone calls with respect to the total call volume. This value must not be < 40%, it is a maximizing criterion. The carrier linkage problem is modeled using a performance matrix data of Table 3. Table 3 shows the performance of carriers, by a mobile operator in Ghana. The first column, labeled criteria (\(C_1\), \(\ldots \), \(C_6\) ) is the column for the six criteria. The second column lists the type of criteria as being a maximizing or minimizing criterion. The alternatives (\(A_1\), \(\ldots \), \(A_6\)) are listed in the row below the header. The rest of the table are the performance measures \(X_{ij}\) which are the scores of the various alternatives under the criteria.
Performance matrix table of carriers (alternatives) against respective criteria
Criteria  Type  Alternatives  Mean  Standard deviation  

\(A_1\)  \(A_2\)  \(A_3\)  \(A_4\)  \(A_5\)  \(A_6\)  
\(C_1\)  Max  100  100  100  87.87  100  100  97.9783  4.952052 
\(C_2\)  Max  26.92  97.61  181.42  183.94  112.39  86.14  114.737  60.1068 
\(C_3\)  Max  22.83  110.35  387.89  311.71  229.85  229.85  215.413  132.3365 
\(C_4\)  Min  117.93  88.45  46.77  59.01  48.9  37.48  66.4233  30.73794 
\(C_5\)  Min  11.9  0.15  0  0  0  0  2.00833  4.846279 
\(C_6\)  Min  31.51  35.36  51.05  45.45  36.4  35.46  39.205  7.418816 
Model formulation and computation
 1Calculation of deviations: The deviations \(d_i(A_k, A_l)\) are obtained through a pairwise comparison of the values of the alternatives on each criterion over all the criteria. It should be noted that the deviations are obtained as below:where \(X_{ik}\) and \(X_{il}\) correspond to performance values of two alternatives \(A_k\) and \(A_l\) on a criterion \(C_i\). Tables 4, 5, 6, 7, 8 and 9 presents all possible deviations \(d_i(A_k,A_l)\) from the pairwise comparison of the alternatives on all the criteria.

\(d_i(A_k,A_l) = \{X_{ik}  X_{il}\}\)—maximizing criteria

\(d_i(A_k,A_l) = \{(X_{ik}  X_{il})\}\)—minimizing criteria,

 2Preference evaluation: We now calculate the intensity measure \(P_i(A_k,A_l)\) of the decision maker’s preference of \(A_k\) over \(A_l\) using the deviations \(d_i(A_k,A_l)\). The preference function to be used is the logistic preference function (Amponsah et al. 2012):where d = deviation, \(\sigma \) = standard deviation, d = \(d_i(A_k,A_l)\) and \(\sigma ^2 = \sigma ^2(i)\) Tables 10, 11, 12, 13, 14 and 15 summarizes the values of \(P_i(A_k,A_l)\) for each criterion \(C_i\).$$\begin{aligned} P(d) = \left\{ \begin{array}{ll} 0 &{} d\le 0\\ \frac{1  exp{\left( \frac{2d^2}{\sigma ^2}\right) }}{1 + exp{\left( \frac{2d^2}{\sigma ^2}\right) }} &{} d>0 \end{array} \right. \end{aligned}$$(14)
 3Aggregate Preference Index: The result to be used in further analysis is obtained by the computation of the aggregate preference index. The aggregate preference index is given by:Table 16 shows the values of \((\pi (A_k,A_l))\) for all six carrier alternatives.$$\begin{aligned} \pi (A_k,A_l) = \sum _{i=1}^n(w_iP_i(A_k,A_l)) \end{aligned}$$(15)
 4Partial ranking: From the aggregate preference indices, the Positive Outranking flow for the carrier alternative \(A_j\) is:and the Negative Outranking flow for the carrier alternative$$\begin{aligned} \phi ^+(A_j) = \frac{1}{n1}\sum _{k=1}^n\pi (A_j,A_k) \end{aligned}$$(16)The values of the positive and negative outranking flows are tabulated in Table 17 In the partial ranking step, the following conditions below should be well noted. \(A_k\) is preferred to \(A_l\) if and only if one of the following three conditions is satisfied:$$\begin{aligned} \phi ^(A_j) = \frac{1}{n1}\sum _{k=1}^n\pi (A_k,A_j) \end{aligned}$$(17)
 (a)
\(\phi ^+(A_k) > \phi ^+(A_l)\) and \(\phi ^(A_k) < \phi ^(A_l)\)
 (b)
\(\phi ^+(A_k) > \phi ^+(A_l)\) and \(\phi ^(A_k) = \phi ^(A_l)\)
 (c)
\(\phi ^+(A_k) = \phi ^+(A_l)\) and \(\phi ^(A_k) < \phi ^(A_l)\)
 (a)
 5Complete ranking: Values of the net flow are calculated asTable 20 presents the calculated values of \(\phi (A_k)\) Preference exists between a pair of alternatives \((A_k,A_l)\) if \(\phi (A_k) \ne \phi (A_l)\) Considering the alternatives \(A_k\), and \(A_l\), alternative \(A_k\) is preferred to alternative \(A_l\) if and only if \(\phi (A_k) > \phi (A_l)\) otherwise \(A_k\) is not preferred to alternative \(A_l\). From the above relations, the table of complete ranking flow is shown in Table 21. Tables 22 and 23 list the total inarrows and outarrows as well as providing ranking for the alternatives. The total number of arrows on each carrier alternative node is counted to be five. Optimal ranking is now achieved since all the carrier alternatives have the required maximum of 5 total arrows. From the graph of complete ranking of Fig. 2, it is observed that all carrier alternate nodes are mutually pairwise linked either through an inarrow or outarrow ave the maximum five linkages which is one less than the total number of nodes. Thus, we have complete ranking with the confirmed order being \(A_1\), \(A_2\) and \(A_6\) \(A_5\) , \(A_4\) and \(A_3\).$$\begin{aligned} \phi (A_k) = \phi ^+(A_k)  \phi ^(A_k) \end{aligned}$$(18)
Deviations \(d_1(A_k,A_l)\) on criterion \(C_1\)
\(A_1\)  \(A_2\)  \(A_3\)  \(A_4\)  \(A_5\)  \(A_6\)  

\(A_1\)  0  0  0  12.13  0  0 
\(A_2\)  0  0  0  12.13  0  0 
\(A_3\)  0  0  0  12.13  0  0 
\(A_4\)  − 12.13  − 12.13  − 12.13  0  − 12.13  − 12.13 
\(A_5\)  0  0  0  12.13  0  0 
\(A_6\)  0  0  0  12.13  0  0 
Deviations \(d_2(A_k,A_l)\) on criterion \(C_2\)
\(A_1\)  \(A_2\)  \(A_3\)  \(A_4\)  \(A_5\)  \(A_6\)  

\(A_1\)  0  − 70.69  − 154.5  − 157.02  − 85.47  − 59.22 
\(A_2\)  70.69  0  − 83.81  − 86.33  − 14.78  11.47 
\(A_3\)  154.5  83.81  0  − 2.52  69.03  95.28 
\(A_4\)  157.02  86.33  2.52  0  71.55  26.25 
\(A_5\)  84.47  14.78  − 69.03  − 71.55  0  26.25 
\(A_6\)  59.22  − 11.47  − 95.28  − 97.8  − 26.25  0 
Deviations \(d_3(A_k,A_l)\) on criterion \(C_3\)
\(A_1\)  \(A_2\)  \(A_3\)  \(A_4\)  \(A_5\)  \(A_6\)  

\(A_1\)  0  − 87.52  − 365.06  − 288.88  − 207.02  − 207.02 
\(A_2\)  87.52  0  − 277.54  − 201.36  − 11.95  − 11.95 
\(A_3\)  365.06  227.54  0  76.18  158.04  158.04 
\(A_4\)  288.88  201.36  − 76.18  0  81.86  81.86 
\(A_5\)  207.02  119.5  − 158.04  − 81.86  0  0 
\(A_6\)  207.02  119.5  − 158.04  − 81.86  0  0 
Deviations \(d_4(A_k,A_l)\) on criterion \(C_4\)
\(A_1\)  \(A_2\)  \(A_3\)  \(A_4\)  \(A_5\)  \(A_6\)  

\(A_1\)  0  − 29.48  − 71.16  − 58.92  − 69.03  − 80.45 
\(A_2\)  29.48  0  − 41.68  − 29.44  − 39.55  − 50.97 
\(A_3\)  71.16  41.68  0  12.24  2.13  − 9.29 
\(A_4\)  58.92  29.44  − 12.24  0  − 10.11  − 21.53 
\(A_5\)  69.03  39.55  − 2.13  10.11  0  − 11.42 
\(A_6\)  80.45  50.97  9.29  21.53  11.42  0 
Deviations \(d_5(A_k,A_l)\) on criterion \(C_5\)
\(A_1\)  \(A_2\)  \(A_3\)  \(A_4\)  \(A_5\)  \(A_6\)  

\(A_1\)  0  − 11.75  − 11.9  − 11.9  − 11.9  − 11.9 
\(A_2\)  11.75  0  − 0.15  − 0.15  − 0.15  − 0.15 
\(A_3\)  11.9  0.15  0  0  0  0 
\(A_4\)  11.9  0.15  0  0  0  0 
\(A_5\)  11.9  0.15  0  0  0  0 
\(A_6\)  11.9  0.15  0  0  0  0 
Deviations \(d_6(A_k,A_l)\) on criterion \(C_6\)
\(A_1\)  \(A_2\)  \(A_3\)  \(A_4\)  \(A_5\)  \(A_6\)  

\(A_1\)  0  − 3.85  − 19.54  − 13.94  − 4.89  − 3.95 
\(A_2\)  3.85  0  − 15.69  − 10.09  − 1.04  − 0.1 
\(A_3\)  19.54  15.69  0  5.6  14.65  15.59 
\(A_4\)  13.94  10.09  − 5.6  0  9.05  9.99 
\(A_5\)  4.89  1.04  0  − 9.05  0  0.94 
\(A_6\)  3.95  0.1  − 15.59  − 9.99  − 0.94  0 
Values of \(P_1(A_k,A_l)\) for criterion \(C_1\)
\(l=1\)  \(l=2\)  \(l=3\)  \(l=4\)  \(l=5\)  \(l=6\)  

\(k=1\)  0  0  0  0.9051  0  0 
\(k=2\)  0  0  0  0.9051  0  0 
\(k=3\)  0  0  0  0.9051  0  0 
\(k=4\)  0  0  0  0  0  0 
\(k=5\)  0  0  0  0.9051  0  0 
\(k=6\)  0  0  0  0.9051  0  0 
Values of \(P_2(A_k,A_l)\) for criterion \(C_2\)
\(l=1\)  \(l=2\)  \(l=3\)  \(l=4\)  \(l=5\)  \(l=6\)  

\(k=1\)  0  0  0  0  0  0 
\(k=2\)  7  0.3326 0  0  0  0  0.0908 
\(k=3\)  0.929  0.4045  0  0 0.3183  0.5568  
\(k=4\)  0.9361  0.4744  0.0004  0  0.3401  0.0477 
\(k=5\)  0.4572  0.0151  0  0  0  0.0477 
\(k=6\)  0.238  0  0  0  0  0 
Values of \(P_3(A_k,A_l)\) for criterion \(C_3\)
\(l=1\)  \(l=2\)  \(l=3\)  \(l=4\)  \(l=5\)  \(l=6\)  

\(k=1\)  0  0  0  0  0  0 
\(k=2\)  0.1089  0  0  0  0  0 
\(k=3\)  0.9565  0.6287  0  0.0827  0.3422  0.3422 
\(k=4\)  0.831  0.5218  0  0  0.0954  0.0954 
\(k=5\)  0.5454  0.2011  0  0  0  0 
\(k=6\)  0.5454  0.2011  0  0  0  0 
Values of \(P_4(A_k,A_l)\) for criterion \(C_4\)
\(l=1\)  \(l=2\)  \(l=3\)  \(l=4\)  \(l=5\)  \(l=6\)  

\(k=1\)  0  0  0  0  0  0 
\(k=2\)  0.226  0  0  0  0  0 
\(k=3\)  0.8716  0.4298  0  0.0396  0.0012  0 
\(k=4\)  0.7252  0.2254  0  0  0  0 
\(k=5\)  0.8513  0.3918  0  0.0271  0  0 
\(k=6\)  0.9369  0.5963  0.0229  0.122  0.0345  0 
Values of \(P_5(A_k,A_l)\) for criterion \(C_5\)
\(l=1\)  \(l=2\)  \(l=3\)  \(l=4\)  \(l=5\)  \(l=6\)  

\(k=1\)  0  0  0  0  0  0 
\(k=2\)  0.8995  0  0  0  0  0 
\(k=3\)  0.9064  0.0002  0  0  0  0 
\(k=4\)  0.9064  0.0002  0  0  0  0 
\(k=5\)  0.9064  0.0002  0  0  0  0 
\(k=6\)  0.9064  0.0002  0  0  0  0 
Values of \(P_6(A_k,A_l)\) for criterion \(C_6\)
\(l=1\)  \(l=2\)  \(l=3\)  \(l=4\)  \(l=5\)  \(l=6\)  

\(k=1\)  0  0  0  0  0  0 
\(k=2\)  0.0672  0  0  0  0  0 
\(k=3\)  0.9396  0.8069  0  0.1398  0.7509  0.802 
\(k=4\)  0.7078  0.432  0  0  0.3557  0.4246 
\(k=5\)  0.1082  0.0049  0  0  0  0.004 
\(k=6\)  0.0707  0.0004  0  0  0  0 
Aggregate Preference Indices \((\pi (A_k,A_l))\)
\(l=1\)  \(l=2\)  \(l=3\)  \(l=4\)  \(l=5\)  \(l=6\)  

\(k=1\)  0  0  0  0.1509  0  0 
\(k=2\)  0.2724  0  0  0.1509  0  0.0151 
\(k=3\)  0.7673  0.3784  0  0.1946  0.2355  0.2836 
\(k=4\)  0.6846  0.2757  0.00007  0  0.1319  0.0946 
\(k=5\)  0.4782  0.1022  0  0.1554  0  0.0086 
\(k=6\)  0.4497  0.133  0.0038  0.1712  0.0058  0 
Values of the positive and negative outranking flows
\(A_j\)  \(\phi ^+(A_j)\)  \(\phi ^(A_j)\) 

\(A_1\)  0.03018  0.53044 
\(A_2\)  0.08768  0.17786 
\(A_3\)  0.37188  0.00077 
\(A_4\)  0.23737  0.1646 
\(A_5\)  0.14888  0.07464 
\(A_6\)  0.1527  0.08038 
Partial ranking flow linkages of carrier alternatives
\(A_1\)  \(A_2\)  \(A_3\)  \(A_4\)  \(A_5\)  \(A_6\)  Outarrows  

\(A_1\)  –  1  1  1  1  1  5 
\(A_2\)  –  –  1  1  1  1  4 
\(A_3\)  –  –  –  –  –  –  0 
\(A_4\)  –  –  1  –  –  –  1 
\(A_5\)  –  –  1  –  –  –  1 
\(A_6\)  –  –  1  –  –  –  1 
InArrows  0  1  5  2  2  2  12 
Partial ranking flow linkages of carrier alternatives
Inarrows  Outarrows  Total  

\(A_1\)  0  5  5 
\(A_2\)  1  4  5 
\(A_3\)  5  0  5 
\(A_4\)  2  1  3 
\(A_5\)  2  1  3 
\(A_6\)  2  1  3 
Values of the net flow for all six alternatives
\(A_k\)  \(\phi ^+(A_k)\)  \(\phi ^(A_k)\)  \(\phi (A_k)\) 

\(A_1\)  0.03018  0.53044  − 0.50026 
\(A_2\)  0.08768  0.17786  − 0.09018 
\(A_3\)  0.37188  0.00077  0.371106 
\(A_4\)  0.23737  0.1646  0.072774 
\(A_5\)  0.14888  0.07464  0.07424 
\(A_6\)  0.1527  0.08038  0.07232 
Complete ranking flow linkages of carrier alternatives
\(A_1\)  \(A_2\)  \(A_3\)  \(A_4\)  \(A_5\)  \(A_6\)  Outarrows  

\(A_1\)  –  1  1  1  1  1  5 
\(A_2\)  –  –  1  1  1  1  4 
\(A_3\)  –  –  –  –  –  –  0 
\(A_4\)  –  –  1  –  –  –  1 
\(A_5\)  –  –  1  1  –  –  2 
\(A_6\)  –  –  1  1  1  –  3 
InArrows  0  1  5  4  3  2  12 
Complete ranking InArrow and OutArrow linkages on carrier alternative nodes
Inarrows  Outarrows  Total  

\(A_1\)  0  5  5 
\(A_2\)  1  4  5 
\(A_3\)  5  0  5 
\(A_4\)  4  1  5 
\(A_5\)  3  2  5 
\(A_6\)  2  3  5 
Ranking of six alternatives using PROMETHEE
Alternatives  Number of outdirected arcs  Ranking position 

\(A_1\)  5  1st 
\(A_2\)  4  2nd 
\(A_3\)  0  6th 
\(A_4\)  1  5th 
\(A_5\)  2  4th 
\(A_6\)  3  3rd 
Discussion
There was situation of incomparability illustrated by no linkages between the carrier alternatives \(A_4\), \(A_5\) and \(A_6\) in the partial raking step of the PROMETHEE method; therefore, the alternatives could not be ranked. The complete ranking step satisfied all the conditions of ranking optimality and all carrier alternatives were appropriately ranked from best to worst in the order, \(A_1\), \(A_2\) and \(A_6\) \(A_5\), \(A_4\) and \(A_3\). Amponsah et al. (2012) who first used the logistic preference function reported general comparability with the performance of the Gaussian preference function even though the logistic preference function performed better than the Gaussian preference function in the test case they used. Using the Gaussian preference function to check our results, it was determined that the complete ranking stage was reached by both preference functions with the ranking order being the same.
Conclusion
This paper is the first test of the logistic preference function introduced in the literature by Amponsah et al. (2012), where the performance of five telecommunication companies were ranked based on data from the National Communication Authority in Ghana. On the other hand, this paper uses PROMETHEE methodology to rank communication carrier alternatives of a telecommunication company. The optimal solution was obtained at the complete ranking stage and carrier alternative VTC1 was the best ranking communication carrier choice. The results of this study will have economic impact with the selection of the best communication transmission carrier. This will introduce efficiency, realtime savings and customer satisfaction. Comparison with results using the Gaussian preference function showed that both approaches reached the same optimal solution at the complete ranking stage.
Declarations
Authors' contributions
BB, JKA and KFD designed the study. JKA and KFD gave technical support and conceptual advice. BB set up the experiment for performance evaluation. All authors prepared the manuscript. All authors read and approved the final manuscript.
Competing interests
The authors declare that they have no competing interests.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Authors’ Affiliations
References
 AlTarawneh HA (2012) The main factors beyond decision making. J Manag Res 4(1):1–23Google Scholar
 Amponsah SK, Darkwah KF, Inusah A (2012) Logistic preference function for preference ranking organization method for enrichment evaluation (PROMETHEE) decision analysis). Afr J Math Comput Sci Res 5(6):112–119View ArticleGoogle Scholar
 Anagnostopolous KP, Petalas C, Pisinaras V (2005) Water resources planning using the AHP and PROMETHEE multicriteria methods: the case of Nestos river—greece. In: The 7th Balkan Conference on Operational Research, Constanta, Romania, 2005Google Scholar
 Anderson R, Mansingh G (2014) Data mining approach to decision support in social welfare. Int J Bus Intell Res 5:39–61View ArticleGoogle Scholar
 Athawale VM, Chakraborty S (2010) Facility location selection using PROMETHEE II method. In: Proceedings of the 2010 international conference on industrial engineering and operations management, Bangladesh, 9–10 January 2010Google Scholar
 Barton H, Beynon MJ (2006a) “Getting to the Top”. Targeted police performance improvement, British Academy of Management (BAM), BelfastGoogle Scholar
 Barton H, Beynon MJ (2006b) A question of rank: an investigation of the opportunities for police force performance ranking improvement across the UK. American Academy of Management, Atlanta, GAGoogle Scholar
 Brans JP, Mareschal B (1992) PROMETHEE V: MCDM problems with segmentation constraints. INFOR: Inf Syst Oper Res 30(2):85–96MATHGoogle Scholar
 Brans JP, Mareschal P (1994) The prometheegaia decision support system for multicriteria investigations. Investig Oper 4(2):107–117Google Scholar
 Brans JP, Mareschal B (1995) The PROMETHEE VI procedure: how to differentiate hard from soft multicriteria problems. J Decis Syst 4(3):213–223View ArticleGoogle Scholar
 Brans JP, Vincke PH (1985) A preference ranking organization method: (the PROMETHEE method for multiple criteria decisionmaking). Manag Sci 31(6):647–656MATHView ArticleGoogle Scholar
 Brans JP, Mareschal B, Vincke P (1986) How to select and how to rank projects: the PROMETHEE method. Eur J Oper Res 24(2):228–238MathSciNetMATHView ArticleGoogle Scholar
 Caron E, Daniels H (2013) Explanatory business analytics in olap. Int J Bus Intell Res 4:67–82View ArticleGoogle Scholar
 Charnes A, Cooper WW (1961) Management models and the industrial applications of linear programming. Wiley, New YorkMATHGoogle Scholar
 Charnes A, Cooper WW, Ferguson RO (1955) Optimal estimation of executive computing by linear programming. Manag Sci 1(2):138–151MATHView ArticleGoogle Scholar
 ChingTer C, ZhengYun Z, HuangMu C (2011) A goal programming modeling for telecommunication routing management. World Acad Sci Eng Technol 51:506–514Google Scholar
 Chromy E, Suran J, Kovacik M, Kavacky M (2011) Usage of Erlang Formula in IP Networks. Commun Netw 3(3):161–167View ArticleGoogle Scholar
 Contini B, Zionts S (1968) Restricted bargaining for organizations with multiple objectives. Econometrica 36:397–414MATHView ArticleGoogle Scholar
 Deshmukh SC (2013) Preference ranking organization method of enrichment evaluation (PROMETHEE). Int J Eng Sci Invent 2(11):28–34Google Scholar
 Diaz R, Behr GJ, Landaeta R, Longo F, Nicoletti L (2015) Modeling energy portfolio scoring: a simulation framework. Int J Bus Anal. https://doi.org/10.4018/IJBAN.2015100101 View ArticleGoogle Scholar
 Figueira JR, Greco S, Roy B (2009) ELECTRE methods with interaction between criteria: an extension of the concordance index. Eur J Oper Res 199:478–495MathSciNetMATHView ArticleGoogle Scholar
 Figueira JR, Greco S, Roy B, Słowiński R (2013) An overview of ELECTRE methods and their recent extensions. J MultiCriteria Decis Anal 20:61–85View ArticleGoogle Scholar
 Hirschberger M, Steuer RE, Utz S, Wimmer M, Qi Y (2013) Computing the nondominated surface in tricriterion portfolio selection. Oper Res 61(1):169–183MathSciNetMATHView ArticleGoogle Scholar
 Ignaccolo M, Inturri P, GarcíaMelón M, Giuffrida N, Pira ML, Torrisi V (2017) Combining analytic hierarchy process (AHP) with roleplaying games for stakeholder engagement in complex transport decisions. Transp Res Procedia 27:500–507View ArticleGoogle Scholar
 Kuhn HW, Tucker AW (1951) Nonlinear programming. In: Proceedings of the second Berkeley symposium on mathematical statistics and probability. University of California Press, Berkeley, pp 481–492Google Scholar
 Kumar S, Parashar N, Haleem A (2009) Analytical hierarchy process applied to vendor selection problem: small scale, medium scale and large scale industries. Bus Intell J 2(2):355–362Google Scholar
 Majumder M, Ghosh S (2013) Decision making methodology. Decision making algorithms for hydropower plant location. Springer Singapore, Singapore, pp 21–26View ArticleGoogle Scholar
 Merkin BG (1979) Group choice. Wiley, New YorkGoogle Scholar
 Moffett A, Sarkar S (2006) Incorporating multiple criteria into the design of conservation area networks: a minireview with recommendations. Divers Distrib 12(2):125–137View ArticleGoogle Scholar
 Nemhauser GL, Kan AHGR, Todd MJ (1989) Handbooks in operations research and management science, vol 1. NorthHolland, AmsterdamGoogle Scholar
 Pearson JM, Ellram LM (1995) Supplier selection and evaluation in small versus large electronics firms. J Small Bus Manag 33(4):53Google Scholar
 Podvezko V, Podviezko A (2010) Dependence of multicriteria evaluation result on choice of preference functions and their parameters. Technol Econ Dev Econ 16(1):143–158View ArticleGoogle Scholar
 Saaty TL (1977) A scaling method for priorities in a hierarchichal structure. J Math Psychol 15:234–281MATHView ArticleGoogle Scholar
 Saaty TL (1980) The analytic hierarchy process. McGrawHill, New YorkMATHGoogle Scholar
 Saaty TL (1986) Axiomatic foundation of the analytic hierarchy process. Manag Sci 32:841–855MathSciNetMATHView ArticleGoogle Scholar
 Saaty TL (1990) An exposition of the AHP in reply to the paper “remarks on the analytic hierarchy process”. Manag Sci 36:259–268View ArticleGoogle Scholar
 Saaty TL (2003) Decisionmaking with the ahp: why is the principal eigenvector necessary?”. Eur J Oper Res 145:85–91MathSciNetMATHView ArticleGoogle Scholar
 Singh RK, Garg SK, Deshmukh SG (2003) Opportunities and challenges for small and medium enterprises in India. In: Proceedings for national conference on recent development in mechanical engineering, Thapar Institute of Engineering and Technology, Patiala, India, 2003Google Scholar
 Singh RK, Garg SK, Deshmukh SG (2005) Development of competences by Indian small, medium and large scale organizations. In: Proceedings of 14th international conference on mechanical engineering in knowledge age, Delhi College of Engineering, Delhi, India, 2005Google Scholar
 Steuer RE (1986) Multiple criteria optimization: theory, computation and application. Wiley, New YorkMATHGoogle Scholar
 Steuer RE, Na P (2003) Multiple criteria decision making combined with finance: a categorized bibliography. Eur J Oper Res 150(3):496–515MATHView ArticleGoogle Scholar
 Tam CYM, Tummala VMR (2001) An application of the AHP in vendor selection of a telecommunications system. Omega 29:171–182View ArticleGoogle Scholar
 Triantaphyllou E, Mann SH (1995) Using the analytic hierarchy process for decision making in engineering applications: some challenges. Int J Ind Eng Theory Appl Pract 2(1):35–44Google Scholar
 Velasquez M, Hester TP (2013) An analysis of multicriteria decision making methods. Int J Oper Res 10(2):56–66MathSciNetGoogle Scholar