Many MCDM methods have been used for technology selection, such as Preference Ranking Organization Method for Enrichment Evaluations (PROMETHEE), method for regeneration technologies assessment and selection (Pohekar and Ramachandran, 2004; Khelifi et al., 2006; Behzadian et al., 2010), Best Available Technology (BAT) framework in multiple industries (Georgopoulou et al., 2007; Liu and Wen, 2012; Chung et al., 2013; Ibáñez-Forés et al., 2013, 2014), the Elimination and Choice Expressing Reality (ELECTRE-II) method in environmental issues (Pohekar and Ramachandran, 2004; Hatami-Marbini et al., 2013; Wen et al., 2015), reference point technique and subjective weight factors methods in technology assessment (Duijm, 2002), expert judgment based models (Shehabuddeen et al., 2006; Daim and Intarode, 2009), TOPSIS in analysis of alternatives for selection of enabling technology (Georgiadis et al., 2013), Multi-Criteria Decision Aiding Hybrid Algorithm (THOR) in waste recycling technology selections (Gomes et al., 2008), VlseKriterijumska Optimizacija IKompromisno Resenje (VIKOR) in platform selection (Lin et al., 2016), Analytical Network Process (ANP) (Molinos-Senante et al., 2015), fuzzy analytic hierarchy process (F-AHP) and fuzzy TOPSIS (Taylan et al., 2016), fuzzy Delphi integrated with fuzzy AHP as well as crisp AHP in lubricant regenerative technology selection (Hsu et al., 2010; Hsu and Liu, 2011), and graphical and traditional methods in waste lubricant regenerative technology selection (Chari et al., 2012).

In addition to the mentioned methods which have been applied in technology selection problems, there are some approaches with hybrid algorithms. For example, Cid-López et al. (2016) suggested a linguistic multi-criteria decision-making model based on 2-tuple linguistic labels for application in ICT services.

Axiomatic Design (AD) principles are developed by Suh (1990) with an ultimate goal of establishing a scientific basis that help designers in logical thought processes (Suh, 1990, 2001; Suh et al., 1998). Kulak and Kahraman (2005b) presented a fuzzy version of AD approach for solving MCDM problems under uncertainty. Significance coefficients of criteria have been added to Fuzzy Axiomatic Design (FAD) model by Kulak et al. (2005) to form the Weighted Fuzzy Axiomatic Design (WFAD) approach.

The FAD and WFAD methods were employed for finding optimal and suitable technology and machine in a manufacturing system. Hierarchical Fuzzy Axiomatic Design (HFAD) approach has been developed by Kahraman and Çebi (2009) to solve a hierarchical teaching assistant selection problem. Gören and Kulak (2014) added risk factors for ratings of alternatives on criteria to HFAD method to generate Risk-Based Fuzzy Axiomatic Design (RFAD) technique. Gören and Kulak employed RFAD approach to specify the best supplier of classic travertine. Kulak et al. (2015) examined the application of RFAD approach in a decision-making problem by considering risks related to criteria regarding the selection of medical imaging devices. Hafezalkotob and Hafezalkotob (2016a) utilized RFAD method with the integrated Shannon entropy significance coefficients. The integration of Shannon entropy with a multi-attribute decision-making method can generate a more robust decision-aid technique. The proposed method was employed in a material selection problem regarding industrial gas turbine blades which are exposed to a shocked temperature and a risk factor is calculated to consider the effect of the difference between the primary and shocked temperature domains on material property.

Many AD applications in the designing systems, products, organization, software, and other interdisciplinary areas appeared in the literature for 27 years. Three detailed literature reviews exist on AD principles and applications. Kulak et al. (2010) reviewed both the crisp and fuzzy AD approaches and provided an overview of the literature on AD principles by covering and classifying 63 papers from 1990 to 2010. Rauch et al. (2016) reviewed the applications of AD in manufacturing system design by covering 20 years of research from 1996–2015. Büyüközkan and Göçer (2017) considered the fuzzy AD principles from 2010 to 2015, by covering 28 papers in the past five years. They combined intuitionistic fuzzy sets with AD methodology in a supplier selection problem. Cheng et al. (2017) presented a novel Heterogeneous Axiomatic Design (HAD) method to solve heterogeneous multi-attribute decision making (HMADM) problems involving deviation criteria, which cannot be solved by ordinary HMADM approaches. The proposed technique was employed in the anti-vibration optimization of the key components in a turbo-generator when a set of alternative schemes are provided.

Kir and Yazgan (2016) proposed a tabu search and genetic algorithm to prepare proper schedules and compute the penalty costs of earliness and tardiness in a scheduling problem of producing milk products on a single machine. This methodology is based on the concept of FAD technique. Integrating the fuzzy Simple Multi-Attribute Rating Technique (SMART) and the WFAD approach, Çakır (2016) employed the proposed methodology to handle a decision-making problem of selecting continuous fluid bed tea dryer. Hou et al. (2016) introduced a new method to analyse and compare alternative schemes based on AD theory and the Markov function that was used to determine the optimization direction of the chosen scheme.

Khandekar and Chakraborty (2015) presented an application of the FAD approach in a material handling equipment selection considering two real-world problems which are selection of an automated guided vehicle, and selection of loading and hauling equipment in surface mines. Chen et al. (2016) suggested a new matching degree calculation method for effectively matching suitable knowledge service demanders and suppliers based on a FAD approach along with a multi-objective optimization model.

Guo et al. (2017) applied the FAD approach in a green supplier evaluation and selection in an apparel manufacturer in Hong Kong. Zheng et al. (2017) developed a rough set FAD approach for performance evaluation and the selection of appropriate additive manufacturing (AM) process.

Chakraborty et al. (2017) proposed a hybrid approach considering FAD and Fuzzy Analytic Hierarchy Process (FAHP) methods in a product design remanufacturing process in India.

Besides the AD priniciples which are target-based approaches based on common area of membership functions of alternative ratings and target values of attributes, there are specific types of normalization methods which allow decision makers to assign targets for criteria. For example, Hafezalkotob and Hafezalkotob (2017) proposed a target-based VIKOR approach based on interval values supported on interval distance and preference degree for two machine selection problems concerning punching equipment and continuous fluid bed tea dryer.

Lubricating oil as a crucial element of hazardous waste in today’s increasing usage of internal combustions engines is one the most valuable liquids that is used in almost all vehicles and industrial machines (Mohammed et al., 2013). Nevertheless, waste oil can be a treasured resource because it contains a significant amount of base oil which may be used to formulate new lubricants by separating undesirable pollutants from the oil by the optimal regenerative process (Rincón et al., 2007). Modern lubricating oils are mixtures of base stock or base oil (71.5–96.2 wt%) blended with an amount of part per million chemical additives to meet the specific requirements of desirable lubricant (Mohammed et al., 2013; Grimes and Thompson, 2016). The recycling of used lubricant oil was in practice to various degrees during the Second World War when the shortage of sufficient supplies of crude oil made a huge constraint for the material consumption and encouraged the reuse of all types of materials including lubricating oils (Abdulkareem et al., 2014; Jafari and Hassanpour, 2015).

Nowadays, there is a serious consideration for sustainability and environmental issues. Three options basically exist to tackle the problem of waste oil: (a) disposing the used oil into water surfaces, land, and sewerage system and garbage heap, (b) regenerating base oil from waste oil, and (c) heat extraction from waste oil through combustion processes (Mohammed et al., 2013). Disposing of the waste oil onto the environment or extraction of heat from the oil will lead to critical environmental consequences. Therefore, because of the great value of waste oil, high possibility of waste recovery and continuous and ongoing care for the sustainability of the green environment, regenerative lubricant oil technologies increasingly come to be seen (Kupareva et al., 2013).

Previous researches (Dalla et al., 2003; Kanokkantapong et al., 2009; Hsu et al., 2010; Chari et al., 2012; Jafari and Hassanpour, 2015) show that many types of lubricant recycling technologies exist that are commercially and technically available and popular such as: distillation process, acid/clay process, Thin Film Evaporation (TFE) with hydro-finishing, TFE with solvent finishing, TFE with clay finishing, solvent extraction hydro-finishing, Thermal De-Asphalting (TDA) with hydro-finishing, and TDA with clay finishing.

Some differences exist among these technologies considering multiple criteria, i.e. economic issues, evolution, and environmental impacts (Hsu et al., 2010). Two comprehensive compendiums of recycling technologies have been developed by United Nations (UN) (Dalla et al., 2003; Chari et al., 2012) which introduce commercial and prototype technologies.

Kupareva et al. (2013) reviewed the regenerative technologies for used lubricant oils applied in Europe and examined 28 plants treating waste oils. Jafari and Hassanpour (2015) presented a comprehensive survey on operational processes of used lubricants regenerative technologies and a comparison of these technologies. There are vast numbers of laboratory researches into the new and novel technologies of lubricant recycling methods that are still in prototype level and not yet provided commercially; i.e. ionic liquid processes (Grimes and Thompson, 2016), various solvent extraction techniques (Al-Zahrani and Putra, 2013).

Only three applications of the RFAD method, i.e. supplier selection problem (Gören and Kulak, 2014), the selection problem of appropriate medical imaging system (Kulak et al., 2015), and material selection problem for industrial gas turbine blade (Hafezalkotob and Hafezalkotob, 2016a) have been presented in the literature. Therefore, this paper presents a new application for the RFAD approach. Regarding to Section 2.1, only four studies have tackled waste lubricant oil regenerative technology selection problems using the MCDM methods, i.e. fuzzy Delphi integrated with fuzzy AHP as well as crisp AHP technique (Hsu et al., 2010; Hsu and Liu, 2011), PROMTHEE approach (Vranes et al., 1999), and graphical and traditional methods (Chari et al., 2012). To the best of authors’ knowledge, no study has considered AD approaches in used lubricant regenerative technology selection. Moreover, no research was conducted on technology selection with consideration of risk factors. However, for uncertain or unexpected situations in some real-world cases like lubricant regenerative technology selection, a risk-based MADM technique is required.

The focus of this paper is to provide a risk-based selection process employing the RFAD approach weighted by the integrated Shannon entropy significance coefficients to find the optimal lubricant regenerative technology among various alternatives based on different criteria. Another contribution of this paper is the proposition of risk factors for the problem of lubricant regenerative technology selection. In this study, after identification of general and specific technology-related risk factors, a comprehensive evaluation for each factor is given.

The decision-making process for the practical case has been classified into two categories. First, by considering all criteria and second, by taking account of the categorized criteria, i.e. the technical and economic aspects. The classification can improve the robustness of the decision process by showing the importance of the economic criteria. There is no study considering categorized criteria in lubricant regenerative technology selection problem.

AD technique is a systematic tool and substitute for traditional MADM methods also helpful for engineers to overlook the whole process of design phase (Khandekar et al., 2015; Hafezalkotob and Hafezalkotob, 2016a). On the other hand, AD principles can evaluate system capabilities by measuring how well the system can satisfy the functional requirements (FRs) (Kulak et al., 2015).

AD principle includes two axioms which are Independence and Information Axioms. Independence Axiom keeps the independence of FRs and Information Axiom keeps the information content at the minimum level (Kulak et al., 2015). Nevertheless, based on the information axiom, the optimal design has the minimum value of the information content. The relationship between the DPs and FRs can be represented as follows: (1) ( FR ) = [ A ] × ( DP ) , in which:

( FR ) denotes the functional requirement vector,

( DP ) shows the design parameter vector, and

[ A ] demonstrates the design matrix which outlines the design.

Referring to Eq. (1), in general, each entry a i j of [ A ] relates the ith F R to the jth D P (Kulak et al., 2010).

Information Axiom presents the information through information content I i j : (2) I i j = log 2 ( 1 p i j ) , in which p i j is the probability of satisfying the ith FR. The reason for using a logarithmic function is that the information content will be additive in the case that several FRs exist that have to be satisfied simultaneously (Kulak et al., 2010). Subsequently, when n FRs exist, the total information content will be the sum of all individual I i j (Fengqiang et al., 2008; Kulak et al., 2015).

In any design situation on the basis of AD principles, the success probability of the design is dependent on two elements; design range, what the designer considers to gain in terms of the expected domain and; system range, what the system is capable of delivering (Kulak et al., 2015; Hafezalkotob and Hafezalkotob, 2016a). The overlap of the two elements (design and system ranges) is named common range, which is the domain of the acceptable design solution (Kulak and Kahraman, 2005a).

As shown in Fig. 1, x-axis illustrates the functional requirement and y-axis denotes the function of probability density. When probability distribution function (PDF) is uniform, p i j is obtained as follows: (3) p i j = ( Common area System area ) .

Hence, the information content I i j is calculated as: (4) I i j = log 2 ( System area Common area ) .

The optimal alternative is the one with the minimum total information content. This feature of the AD method is broadly employed in MADM problems to compare available alternatives (Kulak et al., 2015).

In the real-world decision-making problems, uncertain information is often utilized rather than precise data. The AD principles cannot tackle these practical cases because information is imprecise and there is no probability distribution function available. The fuzzy sets theory approaches could give a solution to human reasoning in uncertain areas. The theory of fuzzy numbers was particularly introduced to deal with vagueness and uncertainty (Zadeh, 1965; Hsu et al., 2010; Gören and Kulak, 2014). Therefore, vague information about the design and system ranges could be presented employing linguistic terms which are then converted into fuzzy numbers (Kulak et al., 2015).

In addition to fuzzy numbers, there are linguistic approaches which could deal with vague information (Liao et al., 2017; Morente-Molinera et al., 2017b). Morente-Molinera et al. (2017a) proposed a multi-granular fuzzy linguistic modelling which allows each expert to choose the linguistic label set. Li et al. (2017) applied a Personalized Individual Semantics (PIS) method to personalize individual semantics by means of an interval numerical scale and the 2-tuple linguistic model in group decision making problem. Capuano et al. (2017) proposed a hybrid model which adopts fuzzy rankings in order to collect both experts preferences on available alternatives and trust statements on other experts which could be useful in situations such as incomplete information availability. Liu et al. (2017) defined the preference relation with self-confidence by taking multiple self-confidence levels into consideration based on heterogeneous preference relations with self-confidence.

In the FAD method, design and system ranges can be represented as trapezoidal or triangular fuzzy sets. In the current study, triangular fuzzy sets are utilized: (5) System range = x ˜ i j = ( x i j , 1 , x i j , 2 , x i j , 3 ) , (6) Design range = d ˜ j = ( d j , 1 , d j , 2 , d j , 3 ) , in which i and j respectively stand for alternatives and criteria, i = 1 , 2 , … , m and j = 1 , 2 , … , n.

As aforementioned, design ranges are ideal properties that are considered by decision-makers on the basis of FRs. Common area c i j is the overlap of system and design areas as demonstrated in Fig. 2. Eq. (7) shows the formula of c i j when the membership functions μ ( x ) of system and ranges areas have similar shape. (7) c i j = If x i j , 2 > d j , 2 , { If x i j , 1 < d j , 3 , ( d j , 3 − x i j , 1 ) 2 2 [ ( d j , 3 − x i j , 1 ) + ( x i j , 2 − d j , 2 ) ] , otherwise , 0 , If x i j , 2 < d j , 2 , { If x i j , 3 > d j , 1 , ( d j , 1 − x i j , 3 ) 2 2 [ − ( d j , 1 − x i j , 3 ) − ( x i j , 2 − d j , 2 ) ] , otherwise , 0 , If x i j , 2 = d j , 2 , common area = system area . {d_{j,2}},\hspace{1em}& \bigg\{\begin{array}{l@{\hskip4.0pt}l}\mathrm{If}\hspace{2.5pt}{x_{ij,1}}<{d_{j,3}},\hspace{1em}& \frac{{({d_{j,3}}-{x_{ij,1}})^{2}}}{2[({d_{j,3}}-{x_{ij,1}})+({x_{ij,2}}-{d_{j,2}})]},\\ {} \mathrm{otherwise},\hspace{1em}& 0,\end{array}\\ {} \mathrm{If}\hspace{2.5pt}{x_{ij,2}}<{d_{j,2}},\hspace{1em}& \bigg\{\begin{array}{l@{\hskip4.0pt}l}\mathrm{If}\hspace{2.5pt}{x_{ij,3}}>{d_{j,1}},\hspace{1em}& \frac{{({d_{j,1}}-{x_{ij,3}})^{2}}}{2[-({d_{j,1}}-{x_{ij,3}})-({x_{ij,2}}-{d_{j,2}})]},\\ {} \mathrm{otherwise},\hspace{1em}& 0,\end{array}\\ {} \mathrm{If}\hspace{2.5pt}{x_{ij,2}}={d_{j,2}},\hspace{1em}& \mathrm{common}\hspace{2.5pt}\mathrm{area}=\mathrm{system}\hspace{2.5pt}\mathrm{area}.\end{array}\right.\]]]>

The probability of satisfying design range d ˜ j for alternative i, i.e. p i j , is obtained by considering common area c i j and system area x ˆ i j as follows: (8) p i j = ( Common area System area ) = ( c i j x i j ˆ ) .

Similar to AD approach, the information content I i j in FAD method is also specified by calculating logarithm with base 2 of the reverse of probability p i j (Gören and Kulak, 2014): (9) I i j = log 2 ( 1 p i j ) .

The total information content is computed for each alternative as follows (Hafezalkotob and Hafezalkotob, 2016a): (10) T i = ∑ j = 1 n I i j .

T i denotes the assessment value of the FAD method. The best alternative on the basis of this technique has the minimum T i , which is demonstrated as follows: (11) A FAD ∗ = { A i | min i T i } .

The importance of criteria in decision-making problems is often not similar. Consequently, to achieve an optimal realistic solution, the relative importance of criteria should be considered. In general, the significance coefficients can be computed using objective, subjective, or integrated techniques (Hafezalkotob and Hafezalkotob, 2016a). Subjective significance coefficients are achieved from the opinions of experts while objective significance coefficients are obtained employing the values of decision matrix without utilizing the judgments of experts. The two types of significance coefficients may be combined. Different techniques exist for calculating significance coefficients of criteria (Rezaei, 2015; Alemi-Ardakani et al., 2016; Hafezalkotob and Hafezalkotob, 2016b, 2016c).

Entropy is based on the classical measures of Boltzmann and the second law of thermodynamics (Zhang et al., 2011; Hafezalkotob and Hafezalkotob, 2016a). The idea of entropy in information science originally suggested by Shannon (1948) is a tool for specifying uncertainty of a variable. The general concept of Shannon’s entropy is to evaluate significance coefficient of each criteria from the distribution of data over variables.

The Shannon entropy has been utilized with combinations of many MCDM tecchniques for various applications; e.g. material selection problems (Hafezalkotob and Hafezalkotob, 2015, 2016a, 2016b, 2016c), competitiveness in tourism industry (Zhang et al., 2011), operational methods for irrigation canals (Shahdany and Roozbahani, 2016), supplier selection problem in petroleum industry facilities (Wood, 2016), phase change materials (Rastogi et al., 2015), and Combined Heat and Power (CHP) systems evaluation (Cavallaro et al., 2016).

The Shannon’s entropy approach also has been combined with many other weighting methods in order to calculate the criteria significance coefficients. Zavadskas and Podvezko (2016) combined the best features of the Shannon entropy method and the criterion impact loss (CILOS) approach to obtain a new method which is Integrated Determination of Objective CRIteria Weights (IDOCRIW). The proposed novel method was combined with FAHP method for application in a contract quality assurance evaluation problem (Trinkūnienė et al., 2017).

The following procedure has to be considered to compute Shannon entropy significance coefficients, assuming that x i j denotes the rating of an alternative on a criterion in a crisp MCDM problem.

Step 1: Normalization of x i j to determine f i j , which is the project outcome (Jahan and Edwards, 2015): (12) f i j = x i j ∑ i = 1 m x i j . Step 2: Calculation of Shannon entropy measure E j using the project outcomes f i j (Hwang and Yoon, 1981): (13) E j = − k ∑ i = 1 m ( f i j ln f i j ) in which k = 1 ln ( m ) . Step 3: Calculation of objective significance coefficients employing E j (Hwang and Yoon, 1981): (14) w j o = d j ∑ j = 1 n d j in which d j = 1 − E j . Step 4: Calculation of the integrated Shannon significance coefficients, if the expert assigns subjective significance coefficients w j s (Hwang and Yoon, 1981): (15) w j ∗ = w j s w j o ∑ j = 1 n w j s w j o .

When w j o , i.e. objective significance coefficient is larger, the variation degree of ratings on the criteria is higher, which is a result of a smaller E j of a criterion. Adversely, larger  E j denotes lower degree of variation of ratings, the less information over criterion j, and minor objective significance coefficient w j (Hafezalkotob and Hafezalkotob, 2016a).

To use the integrated Shannon entropy significance coefficients in FAD approach, the fuzzy ratings of alternatives on criteria first have to be defuzzified. The defuzzified value of system range x ˜ i j for triangular fuzzy sets can be calculated as follows (Allahviranloo and Saneifard, 2012; Ahmed et al., 2014): (16) x ‾ i j = x i j , 1 + 4 x i j , 2 + x i j , 3 6 .

Based on the aforementioned steps of calculating the integrated Shannon significance coefficients and the defuzzified system ranges x ‾ i j , the integrated Shannon significance coefficients are determined. The integrated significance coefficients are then used to generate the information content of the WFAD ∗ approach, i.e. I i j w ∗ , as follows: (17) I i j w = ( I i j ) 1 / w j ∗ , 0 ⩽ I i j ⩽ 1 , ( I i j ) w j ∗ , I i j > 1 , w j ∗ , I i j = 1 . 1,\\ {} {w_{j}^{\ast }},\hspace{1em}& {I_{ij}}=1.\end{array}\right.\]]]>

The total information content and the optimal alternative based on the WFAD method are computed as: (18) T i w = ∑ j = 1 n I i j w , (19) A WFAD ∗ = { A i | min i T i w } .

Risk can be defined differently based on its application. “The possibility of loss, injury, disadvantage or destruction”, as defined in Webster dictionary, is one of the definitions of risk. British standards describe risk as “a combination of the probability of frequency of occurrence of a defined hazard and the magnitude of the consequences of the occurrence” (Subramanyan et al., 2012). Risks are considered as the probability of occurrence of some uncertain, unpredictable, and even undesirable outcome(s) arising from a decision and a negative (or positive) possibility for the profitability of a given investment (Kartam and Kartam, 2001; Okmen and Oztas, 2008; Sharma and Swain, 2015).

By utilizing FAD method with risk factors achieving a practical decision is more likely. Thus, by combining elements of risk and FAD approach, the RFAD method is obtained. The information content for the RFAD technique is calculated as follows Gören and Kulak (2014): (20) I i j r = log 2 ( 1 P i j ( 1 − r i j ) ) , in which r i j is a risk factor with a value in the range of zero and one. Comparing information contents of FAD and RFAD approaches, it is explicit that each I i j r is greater than its corresponding I i j . Greater risk factor r i j leads to higher value of information content I i j r . The total information content and the optimal alternative on the basis of the RFAD approach is obtained as follows: (21) T i r = ∑ j = 1 n I i j r , (22) A RFAD ∗ = { A i | min i T i r } .

Hafezalkotob and Hafezalkotob (2016a) developed the RFAD method with the integrated Shannon entropy significance coefficients to generate entropy-weighted risk-based fuzzy axiomatic design (WRFAD) approach. The information content of the WRFAD technique based on the integrated Shannon significance coefficients is determined as follows: (23) I i j w r = ( I i j r ) 1 / w j ∗ , 0 ⩽ I i j r ⩽ 1 , ( I i j r ) w j ∗ , I i j r > 1 , w j ∗ , I i j r = 1 , 1,\\ {} {w_{j}^{\ast }},\hspace{1em}& {I_{ij}^{r}}=1,\end{array}\right.\]]]> in which w j ∗ denotes the integrated Shannon entropy significance coefficients of criterion j. The total information content and the optimal alternative on the basis of the WRFAD is computed as: (24) T i w r = ∑ j = 1 n I i j w r , (25) A WRFAD ∗ = { A i | min i T i w r } .

In green technology selection process, r i j which denotes the risk factor can be obtained based on comments of decision-makers or by performing specific experiments. The risk factors may be caused from multiple sources such as economic, environmental, and sustainability aspects. In Section 4, Tables 2 and 3 offer a comprehensive description on the general and specific risk factors concerning waste lubricant oil regenerative technologies, respectively.

The process of the proposed WRFAD approach for green technology selection is illustrated in Fig. 3. Design ranges are the considered limits for criteria and system ranges are the performances of the candidate technologies on the criteria. In a typical green technology selection problem, system and design ranges as well as risk factors can be obtained by laboratory experiments, simulations, or based on comments of experts or utilizing handbooks and standards. Significance coefficients for criteria are calculated as described in Section 3.3. The information contents can be computed for the FAD, WFAD, RFAD, and WRFAD methods as explained in Section 3 employing the significance coefficients, probability formula, and risk factors. Eventually, the best green technology and the rankings are determined based on the total information contents.