Indonesian J our nal of Electrical Engineering and Computer Science V ol. 17, No. 1, January 2020, pp. 324 330 ISSN: 2502-4752, DOI: 10.11591/ijeecs.v17i1.pp324-330 r 324 A fuzzy based v ertical hando v er netw ork selection scheme Meenakshi Subramani, V inoth Bab u K umara v elu School of Electronics Engineering, V ellore Institute of T echnology , V ellore, T amil Nadu, India Article Inf o Article history: Recei v ed Apr 1, 2019 Re vised Jun 17, 2019 Accepted Jul 7, 2019 K eyw ords: Analytic hierarch y process (AHP) De vice-to-De vice (D2D) communication Fuzzy logic Hando v er decision delay Simple additi v e weighting (SA W) V ertical Hando v er (VHO) ABSTRA CT One of the most attracti v e and challenging areas in the upcoming ne xt-generation 5G wireless netw ork is the v ertical hando v er (VHO). Recently , man y of the heterogeneous wireless communication technologies are introduced to satisfy the demands of users in all situations. Due to the deplo yment of heterogeneous netw orks, the users can ac- cess the internet an ywhere, an ytime through dif ferent wireless netw orks. T o obtain seamless service and service continuity , the de vice should be handed o v er to the best wireless netw orks. Here, a half hando v er scheme for De vice-to-De vice (D2D) com- munication is implemented for the selection of the best netw ork. The tar get netw ork selection for v ertical hando v er can be handled using multiple attrib ute decision mak- ing (MADM) methods. An intelligent and f ast v ertical hando v er decision is much needed, which should be reliable e v en for random and uncertain en vironments. Fuzzy logic is pro v ed to be ef fecti v e in handling imprecise data. Hence, in this w ork, the impact of combining fuzzy with the con v entional MADM scheme, simple additi v e weighting (SA W) is analyzed and the h ybrid scheme is compared with the con v entional MADM schemes lik e SA W , T echniques for order preference by similarity to ideal so- lution (T OPSIS), VlseKriterijumska optimizacija I K ompromisno Resenje (VIK OR) in terms of hando v er decision delay . Since, the numbers of hando v ers e x ecuted are lo w , the hando v er decision delay performance of the proposed scheme is superior than the considered classical MADM schemes. Copyright c 2020 Insitute of Advanced Engineeering and Science . All rights r eserved. Corresponding A uthor: V inoth Bab u K umara v elu, School of Electronics Engineering, V ellore Institute of T echnology , V ellore, T amil Nadu, India. Email: vinothbab@gmail.com 1. INTR ODUCTION Due to increased demands of the users, 5G will o v ercome the shortages of 4G communication, which is e xpected to of fer higher data rate, incr eased v oice quality calls, impro v ed spectrum ef ficienc y , reduced la- tenc y , etc. to the end users [1]. 5G netw orks will support the emer ging and e xisting technologies, which inte grates ne wer solutions to obtain the increasing demand for higher data rate. One of the important features of 5G is D2D communication [2]. In D2D, tw o de vices in close proximity communicate directly rather than communicating through the e v olv ed node base station (eNB) or the core netw ork. Due to its direct link communication, it of floads data traf fic and impro v es spectral ef ficienc y . It also reduces po wer consumption and latenc y . In 5G, the concept of small cells is v ery popular , where the co v erage range of access points are reduced [3]. Due to the mobility nature, the de vices under going D2D communica- tion requires frequent hando v ers [2]. The selection of the tar get netw ork is to ensure proper communication and seamless service between the de vices, which acts as a basic requirement in man y practical applications. T o achie v e al w ays best connection (ABC) and to meet the good quality of service (QoS), the de vice has to J ournal homepage: http://ijeecs.iaescor e .com/inde x.php/IJEECS f or d e vice - to-de vi c e comm unicatio n Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 r 325 switch to dif ferent netw orks [4]. The process of switching to dif ferent netw orks is called VHO and the process of switching between the same netw orks is called horizontal hando v er [2]. The interesting and challenging research area in the wireless en vironment is to select the best netw ork from se v eral candidate netw orks. When one of the de vices under going D2D communication mo v es a w ay from the other de vice, the link quality be- tween them becomes v ery poor . This leads to poor QoS and connection breakdo wn. Hence, one of the de vices is handed o v er to the other netw ork. After the hando v er , these de vices may continue their communication through cellular links. This type of hando v er is termed as half hando v er . A sample D2D scenario requiring hando v er is sho wn in Figure 1a. D1 and D2 mark ed in Figure 1a represent the de vices under going D2D com- munication. The process of half hando v er is illustrated in Figure 1b . In some cases, both the de vices under going D2D communication may mo v e to w ards the neighboring netw ork. F or seamless service continuity , both the de vices are jointly handed o v er to the best neighboring netw ork. This process is termed as j oint hando v er [2], which is illustrated in Figure 1c. (a) (b) (c) Figure 1. (a) A scenario for the de vice in D2D communication requiring hando v er , (b) After half hando v er , (c) After joint hando v er T ar get netw ork selection is an important task to achie v e a seamless connection and QoS in VHO en vironment. The netw ork g athers the parameters from e v ery candidate netw ork and ranks them to choose the best tar get netw ork. Most of the con v entional hando v er schemes mak e use of recei v ed signal strength (RSS) to select the tar get netw ork [4, 5]. This introduces a ping-pong ef f ect, which leads to decreased throughput, increased latenc y and dropping rate. Man y of the recent approaches mak e use of v arious parameters to decide the hando v er and the tar get netw ork. These include a signal to noise ratio (SNR), achie v able bit rate, bit error rate (BER), outage probability , cost, security , po wer consumption, a v ailable bandwidth, etc. [4]. Due to its simplicity in operation, MADM algorithms are widely used in VHO decision making and tar get netw ork selection. In literature, there e xist man y compensatory and non-compensatory MADM algo- rithms. Most of the netw ork selection algorithms proposed are based on the compensatory method. The compensatory algorithms combine multiple criteria to find the best netw ork, whereas non-compensatory algo- rithms combi ne multiple criteria to find the acceptable netw ork, which satisfies the minimum requirements. SA W [6], T OPSIS [7], VIK OR [8] algorithms come under compensatory cate gory . These are popular for lo wer computational comple xity and impro v ed accurac y in decision making. Fuzzy logic models comple x systems f airly without an y bias, which is not in the case of the AHP process [9, 10]. Fuzzy logic is capable of processing a lar ge number of inputs and mak es a soft decision. It ef ficiently handles the imprecise data and represents it in an innate form [3]. Most of the con v entional VHO decision-making algorithms are based on RSS, which fluctuates based on v arious parameters lik e distance, mobility , speed and shado wing f actor , etc. The imprecise input may cause inaccurate decision in deciding the hando v er . This leads to o v er or under -utilization of netw ork resources. Fuzzy logic is pro v ed to be ef ficient in handling the data related to radio, QoS and user preferences [4]. 2. THE PR OPOSED METHOD In this w ork, fuzzy logic is combined with the con v entional AHP and SA W methods to sim ultaneously process a lar ge number of inputs and to ef fecti v ely handle the imprecise data related to hando v er decision making and tar get netw ork selection. Based on the input criteria, Fuzzy AHP is used to calculate the weights of each criteria. These weights and the a v ailable c andidate netw orks are gi v en as the input for the Fuzzy SA W scheme. The netw ork with the highest rank is chosen as the tar get netw ork to hando v er . The proposed method for multi-criteria netw ork selection block diagram is sho wn in Figure 2. A fuzzy based vertical hando ver network selection... (Meenakshi Subr amani) Evaluation Warning : The document was created with Spire.PDF for Python.
326 r ISSN: 2502-4752 The rest of the w ork is arranged as follo ws: An introduction to fuzzy theory , Fuzzy AHP and Fuzzy SA W are e xplained in section 3. The results are discussed in section 4 and the paper is concluded in section 5. Figure 2. Block diagram of Fuzzy AHP-Fuzzy SA W MADM netw ork selection 3. RESEARCH METHOD 3.1. Fuzzy theory Fuzzy set theory denotes the ambiguous data in an innate form [3]. Due to the lo wer com p ut ational comple xity and simpler mathematical implementation, triangular fuzzy number (TFN) is widely preferred for the applications related to wireless communication. It is also pro v ed that the comple xities related to handling imprecise and uncertainty information used for netw ork selection is minimized with TFN. Hence, in this w ork, TFN is utilized to establ ish ambiguity metrics. TFN is one of the important classification of fuzzy number with its membership function defined by ( x; y ; z ) , three crisp numbers as ( x y z ) , where x is the lo wer limit, y is the modal v alue and z is the upper limit. When x = y = z , then the fuzzy number will become a real number . A fuzzy set e G in a uni v erse of discourse A is represented by a membership function e G ( a ) , which is related to each element a in A and the real number with an interv al [0 ; 1] . The e G ( a ) function v alue is referred as the membership grade of a in e G . The fuzzy number e G on R to be TFN, when the membership function, e G ( a ) : R ! [0 ; 1] . TFN can be defined as e G ( a ) = 8 > < > : ( a x ) ( y x ) ; x a y ( z a ) ( z y ) ; y a z 0 ; other w ise (1) 3.2. Fuzzy AHP AHP is mainly used for decision-making in multi-criteria problems, which is initiated with a measure- ment of ratio scales [11]. AHP use comparison and pre-set option selection, which depends on the pairwise comparisons between the options and criteria. Here, a qualitati v e judgment is used for pairwise comparison. Fuzzy AHP is v ery easy and simple to analyze in making decisions. Fuzzy AHP causes the significance of fuzzy , which occurs in the same ro w with tw o combination items. The fuzzy e xtension is needed because the basic AHP f ailed to address the main issue of handling the high de gree of imprecise in subjecti v e personal judgments and i ts preferences. Fuzzy AHP is applic able to tackle the problem at hand by considering the structure of multi-criteria and v ague- ness in a real-time en vironment, which impro v es the consistenc y in the netw ork selection. Fuzzy AHP uses a 9-point fundamental scale, which is e xpressed in terms of TFN to indicate the relati v e performance between the pairwise decision f actors. The TFN v alues are represented in T able 1. T able 1. 9-point fundamental fuzzy scale using TFN [12] Fuzzy scale intensity of importance Linguistic scale description TFN Reciprocal of TFN e 1 Equal importance (1,1,1) (1,1,1) e 2 Moderately equal important (1,2,3) (1/3,1/2,1) e 3 Moderately important (1,3,5) (1/5,1/3,1) e 4 Moderately strongly important (2,4,6) (1/6,1/4,1/2) e 5 Strongly important (3,5,7) (1/7,1/5,1/3) e 6 Strongly v ery important (4,6,8) (1/8,1/6,1/4) e 7 V ery strongly important (5,7,9) (1/9,1/7,1/5) e 8 V ery strongly e xtremely important (6,8,9) (1/9,1/8,1/6) e 9 Extremely important (7,9,9) (1/9,1/9,1/7) Indonesian J Elec Eng & Comp Sci, V ol. 17, No. 1, January 2020 : 324 330 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 r 327 The Fuzzy AHP has the follo wing steps: Step 1: Based on the objecti v es, select the suitable criteria. Step 2: Construct a pairwise comparison matrix e R based on each criteria. The elements f r ij in e R represents the pairwise comparison of criteria i with j . e R = [ f r ij ] n n = 2 6 6 6 4 f r 11 f r 12 f r 13 : : : f r 1 n f r 21 . . . f r 22 . . . f r 23 r 2 n . . . . . . f r n 1 f r n 2 f r n 3 : : : g r nn 3 7 7 7 5 (2) where n denotes the number of criteria decision compared and i = j = 1 ; 2 ; ::::n: f r ij represents the relati v e strength of tw o elements based on TFN. f r j i = [ f r ij ] 1 = ( x ij ; y ij ; z ij ) 1 = 1 z ij ; 1 y ij ; 1 x ij (3) Step 3: The Fuzzy AHP comparison matrix is represented with TFN as e R = [ f r ij ] n n = 2 6 6 6 4 (1 ; 1 ; 1) ( x 12 ; y 12 ; z 12 ) ( x 1 n ; y 1 n ; z 1 n ) ( x 21 ; y 21 ; z 21 ) . . . (1 ; 1 ; 1) . . . ( x 2 n ; y 2 n ; z 2 n ) . . . ( x n 1 ; y n 1 ; z n 1 ) ( x n 2 ; y n 2 ; z n 2 ) (1 ; 1 ; 1) 3 7 7 7 5 (4) Step 4: The fuzzy geometric mean and fuzzy weight for each criteria is calculated as [13, 14] e l i = ( f r i 1 f r i 2 ::::: f r in ) 1 =n (5) f w i = e l i e l 1 l 2 l n 1 (6) The sign indicates fuzzy multiplication and sign indicates fuzzy addition. f r in is the fuzzy comparison v alue of i th criteria to n th criteria, e l i is the geometric mean of fuzzy comparison v alue of i th criteria to each criteria, f w i is the fuzzy weight of i th criteria. Step 5: In order to check the inconsistenc y in the pairwise comparison matrix, consistenc y inde x (CI) is introduced to obtain the consistenc y ratio (CR). CI is gi v en as C I = max n n 1 (7) where max = e R : e w e w (8) where max is the lar gest Eigen v alue of the comparison matrix e R , n represents the number of criteria and e w is the weight v alue calculated to obtain Eigen v ector . Based on the CI, the v alue of CR is calculated as C R = C I R I (9) where RI is the random consistenc y inde x as sho wn in T able 2. V arious authors measured RI v alues for a dif ferent number of criteria. This is tab ulated in [15]. In [15], the authors estimated RI for each number of criteria using 100,000 matrices. The y ha v e generated random matrices with uniform distrib ution. Then, CIs are calculated for each matrix. The RI for each number of criteria is obtained by taking mean of these CI v alues. It is pro v ed that the RI v alue calculated by [16] is ef ficient than t he other methods. Hence, in this w ork, we ha v e used the same table, which w as listed in [16]. If the estimated CR v alue is less than 0.1, then the pairwise construction is acceptable, otherwise, the matrix has to be re vised [17]. T able 2. Random consistenc y v alue [16] Criteria 1 2 3 4 5 6 7 8 9 10 RI 0 0 0.58 0.9 1.12 1.24 1.32 1.41 1.45 1.49 A fuzzy based vertical hando ver network selection... (Meenakshi Subr amani) Evaluation Warning : The document was created with Spire.PDF for Python.
328 r ISSN: 2502-4752 3.3. Fuzzy SA W Because of its simpli city , Fuzzy SA W is widely used in the MADM algorithms[18, 4]. Fuzzy SA W requires the normalizing procedure for the decision matrix into a scale, which is compared with e v ery netw ork rating. The main idea of Fuzzy SA W method is to detect the weighted sum of performance rating of e v ery netw ork on each criteria. The steps are as follo ws: Step 1: W eight calculation Obtain the weight v alue for each criteria from Fuzzy AHP method. Step 2: Fuzzy decision matrix F ormulate the decision matrix e F and select the suitable linguistic v ariables with respect to dif ferent netw orks and criteria. A q n fuzzy decision matrix is formulated with the ratings of each netw ork with each criteria. The entries are TFN instead of crisp v alues. C 1 C 2 C 3 C n e F = N w 1 N w 2 N w 3 . . . N w q 2 6 6 6 6 6 6 4 f b 11 f b 12 f b 13 f b 1 n f b 21 f b 22 f b 23 f b 2 n f b 31 f b 32 f b 33 f b 3 n . . . . . . . . . . . . f b q 1 f b q 2 f b q 3 f b q n 3 7 7 7 7 7 7 5 (10) where N w 1 ; N w 2 ; N w 3 ; ::::::::: ::N w q are the possible netw orks, C 1 ; C 2 ; C 3 ; ::::::::: ::; C n are the criteria. The element f b ij is the fuzzy decision matrix, which indicates the performance rating of each netw ork N w i with respect to criteria C j , where i = 1 ; 2 ; 3 ; :::::: q and j = 1 ; 2 ; 3 ; :::::n respecti v ely . Step 3: Fuzzy normalization weight v alue The fuzzy decision matrix e F depends on the l inguistic v ariables and the corresponding TFN. Each entry of e F is TFN, which corresponds to 3 v alues, i.e, g b ij x ; g b ij y ; g b ij z : .The normalized fuzzy decision matrix f g ij is gi v en by f g ij =   g b ij x e b j + ; g b ij y e b j + ; g b ij z e b j + ! (11) F or benefit criteria, the maximum of 3rd v alue of TFN in the fuzzy decision matrix e F is identified using e b j + = max i g b ij z ; w her e j 2 B (12) where B is the set of benefit criteria, which is al w ays e xpected to be maximum. All the elements of e F are normalized using e b j + : . Alternati v ely , cost criteria can also be used. The normalized fuzzy decision matrix f g ij for cost criteria is gi v en by f g ij =   e c j g b ij z ; e c j g b ij y ; e c j g b ij x ! (13) F or cost criteria, the minimum of the 1st v alue of TFN in the fuzzy decision matrix e F is identified using e c j = min i g b ij x ; w her e j 2 C (14) where C is the set of cost criteria, which is al w ays e xpected to be minimum. All the elements of e F are normalized using e c j . F or benefit criteria, the highly acceptable v alue is considered as the best v alue. F or e xample, RSS and data rate. F or cost criteria, the lo wer acceptable v alue is considered as the best v alue. Step 4: W eight normalization v alue In this step, a weight normalized decision matrix g M ij is formulated by multiplying each element f g ij of normalized fuzzy decision matrix with the weights e w obtained through Fuzzy AHP in step 4 of section 3 as g M ij = 2 6 6 6 4 f w 1 f g 11 f w 2 f g 12 f w 3 f g 13 f w n f g 1 n f w 1 f g 21 f w 2 f g 22 f w 3 f g 23 f w n f g 2 n . . . . . . . . . . . . . . . f w 1 f g q 1 f w 2 f g q 2 f w 3 f g q 3 f w n f g q n 3 7 7 7 5 (15) Indonesian J Elec Eng & Comp Sci, V ol. 17, No. 1, January 2020 : 324 330 Evaluation Warning : The document was created with Spire.PDF for Python.
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752 r 329 Step 5: Netw ork ranking Fuzzy SA W algorithm e v aluates all the netw orks and mak es a decision to perform hando v er . The o v erall rank of i th netw ork is e v aluated using D S AW i = 1 n n X j =1 f g ij : f w j (16) 4. RESUL TS AND DISCUSSION MA TLAB 2017a tool is used for simulation. The consi dered simulation parameters are li sted in T able 3. Figure 3 sho ws the hando v er decision delay v ersus a number of inputs. The simulations are carried out for streaming applications. During the v ertical hando v er decision, the hando v er decision delay is considered as one of the important parameters. The delay in the hando v er decision leads to QoS de gradation. If the hando v er decision is done too early , it leads to unnecessary hando v er . The goal is to measure the hando v er decision delay for the increased number of inputs. In man y real-w orld problems, decision-mak ers may not be sure about their preferences. This leads to uncertainty . The inclusion of fuzzy logic ef fecti v ely handles v arious decision-making problems. In Fuzzy AHP , the pairwise comparison matrix is formulated with the help of TFN. This a v oids ambiguities in finding the weights. These weights are used in fuzzy SA W for ranking and selecting the tar get netw orks. The decision mak ers also use the fuzzy logic in formulating the decision matrix which compares alternati v es with criteria. Since, the uncertainties are handled ef fecti v ely and f ast, the hando v er decision t ime during the selection of the tar get netw ork is v ery lo w for the Fuzzy AHP-Fuzzy SA W scheme. T able 3. Simulation parameters P arameters V alues Netw orks W i-Fi = W iMAX = L TE-A Cell radius (km) W i-Fi: 0.25; W iMAX: 10; L TE-A: 3 T ransmit Po wer (dBm) W i-Fi: 13; W iMAX: 47; L TE-A: 46 Bandwidth (MHz) W i-Fi: 20; W iMAX: 40; L TE-A: 100 P ath loss model for W i-Fi P L ( dB ) W i F i = 34 : 48 + 32 : 79log 10 d ( m ) [2] P ath loss model for W iMAX P L ( dB ) W iM AX = 130 : 62 + 37 : 6log 10 d ( k m ) [2] P ath loss model for L TE-A P L ( dB ) LT E A = 103 : 8 + 20 : 9log 10 d ( k m ) [2] Figure 3. Hando v er decision delay v ersus number of inputs comparison of v arious MADM schemes F or three criteria, Fuzzy AHP-Fuz zy SA W of fers 29.03 % , 43.59 % , 55.55 % , 20 % , reduction in hando v er decision delay o v er the con v entional SA W [4, 6], T OPSIS [4, 7], VIK OR [4, 8] and Fuzzy SA W [4] schemes. F or an y number of increased inputs, Fuzzy AHP-Fuzzy SA W hando v er decision time will be lesser . Hence, Fuzzy AHP-Fuzzy SA W selects the best tar get netw ork to perform an ef ficient, f ast and seamless hando v er . 5. CONCLUSION Since D2D communication mostly happens for a smaller duration, the hando v er decision delay should be much smaller for a seamless connection. M ost of the con v entional MADM schemes are not f ast and reliable. The y also f ail when handling the imprecise data. Hence, i n t his w ork, the concept of fuzzy is combined with A fuzzy based vertical hando ver network selection... (Meenakshi Subr amani) Evaluation Warning : The document was created with Spire.PDF for Python.
330 r ISSN: 2502-4752 the con v entional SA W and the h ybrid scheme Fuz zy AHP-Fuzzy SA W is pro v ed to of fer impro v ed hando v er decision delay performance o v er SA W , T OPSIS, VIK OR, Fuzzy SA W schemes. Due to lo wer hando v er deci- sion delay , the pack et drop ratio and service interruption are greatly reduced for the proposed scheme. This increases the a v erage data rate e xperienced by each de vice. This also minimizes the latenc y of the proposed scheme. This simple w ork can be e xtended for other criteria, netw orks and MADM algorithms. As a future study , these schemes can also be combined and tested with neuro-fuzzy based algorithms. REFERENCES [1] S.T . Shah, et al., “De vice-to-de vice communications: A contemporary surv e y”, W ireless Personal Com- munications , 98, No.1, pp.1247-1284, 2018. [2] M. Subramani, and V .B. K umara v elu, A Quality-A w are Fuzzy-Logic-Based V ertical Hando v er Decision Algorithm for De vice-to-De vice Communication”, Arabian Journal for Science and Engineering , pp.1-13, 2018. [3] A. Murug adass, and A. P achiyappan, “Fuzzy Logic Based Co v erage and Cost-Ef fecti v e Placement of Serving Nodes for 4G and Be yond Cellular Netw orks”, W ireless Communications and Mobile Comput- ing , 2017. [4] A.B Zineb, et al., An enhanced v ertical hando v er based on fuzzy inference MADM approach for hetero- geneous netw orks”, Arabian Journal for Science and Engineering , 42, No.8, pp.3263-3274, 2017. [5] T . Thumtha w atw orn, et al., Adapti v e Multi-fuzzy Engines for Hando v er Decision in Heterogeneous W ireless Netw orks”, W ireless Personal Communications , 93, No.4, pp.1005-1026, 2017. [6] P . Bagg a, A. Joshi and R. Hans, ”QoS based W eb Service Selection and Multi-Criteria Decision Making Methods”, International Journal of Interacti v e Multimedia and Artificial Intelligence , V ol. 5, No. 4, 2019. [7] H. Lam W eng, S. Lam W eng and F . Lie w Kah, ”Performa n c e analysis on telecommunication companies in Malaysia with T OPSIS model”, Indonesian Journal of Electrical Engineering and Computer Science , V ol. 13, No. 2, PP . 744-751, 2019. [8] S.A.A. Alrababah, et al., “Comparati v e analysis of MCDM methods for product aspect ranking: T OPSIS and VIK OR”, IEEE International Conference on Information and Communication Systems , pp. 76-81, 2017. [9] N. Hamidy , H. Alipur , S.N.H. Nasab, A. Y azdani and S. Shojaei, ”Spatial e v aluat ion of appropriate areas t o collect runof f using Analytic Hierarch y Process (AHP) and Geographical Information System (GIS)(case study: the catchm ent “Kasef” in Bardaskan)” Modeling Earth Systems and En vironment , 2, No. 4, pp.1-11, 2016. [10] S. Shojaei, H. Alipur , A.H.H. Ardakani, S.N.H. Nasab and H. Khosra vi, ”Locating Astrag alus h ypsogeton Bunge appropriate site using AHP and GIS”, Spatial Information Research , 26, No. 2, pp.223-231, 2018. [11] S.A. Mohammed, ”Solving Course Selection Problem by a Combination of Correlation Analysis and Analytic Hier arch y Process”, International Journal of Electrical and Computer Engineering , V ol. 7, No. 6, pp. 3536-3551, 2017. [12] B.M. Elomda, et al., An e xtension of fuzzy decision maps for multi-criteria decision-making”, Egyptian Informatics Journal , 14, No.2, pp.147-155, 2013. [13] M. Chandrashekhar , A. Sh yam Sundar , ”Fuzzy Multi-criteria Decision Making associated with Ris k and Confidence Attrib utes”, Bulletin of Electrical Engineering and Informatics , V ol. 4, No. 3,pp. 231-240, 2015. [14] T .Y . Hsieh, et al., “Fuzzy MCDM approach for planning and design tenders selection in public of fice b uildings”, International journal of project management , 22, No.7, pp.573-584, 2004. [15] J.A. Alonso, and M.T . Lamata, “Consistenc y in the analytic hierarch y process: a ne w approach”, Interna- tional journal of uncertainty , fuzziness and kno wledge-based systems , 14, No.4, pp.445-459, 2006. [16] J.A. Alonso, and M.T . Lamata, “Estimation of the random inde x in the analytic hierarch y process”, In- formation processing and management of uncertainty in kno wledge-based systems , V ol. 1, pp. 317-322, 2004. [17] A. Habbal, et al., “Conte xt-a w are radio access technology se lection in 5G ultra-dense netw orks”, IEEE access , 5, pp.6636-6648, 2017. [18] E. Roszk o wska and D. Kacprzak, “The fuzzy SA W and fuzzy T OPSIS procedures based on ordered fuzzy numbers”, Information Sciences , 369, pp.564-584, 2016. Indonesian J Elec Eng & Comp Sci, V ol. 17, No. 1, January 2020 : 324 330 Evaluation Warning : The document was created with Spire.PDF for Python.