The criteria for the cluster selection for single hop and multi-hop based sensor-cloud environment

Abstract

The major drawbacks of the routing model energy consumption of cluster head (CH) are increased due to their additional functions. Most CH formations depend on a single criterion, which is that CH is generally elected on the basis of randomly or residual energy, sink distance or node density. If the head node is chosen on the basis of residual energy, but the higher energy node is too far away to sink, then the problem appears for election on CH. Similarly, the same problem will arise when distance is shorter to the sink and CH is created near the BS. So, a single criterion is not enough for the chosen head node. A multiple criterion is proposed to overcome this problem. In this paper multi-hopping communication is considered in the cases of intra-cluster and inter-cluster communication, where the CH formation is calculated using Analytical Hierarchy Process (AHP). The outcomes are recorded with 5% improvement when compared to conventional methods in terms of energy efficiency, network life span, control overhead, less cluster head deformation.

Keywords

Analytical hierarchy process intra-cluster communication inter-cluster communication multi-hopping model energy efficiency

1. Introduction

Sensor-clouds play a significant role within the upcoming wireless communication domain due to its properties, viz. intelligence, low cost and small size [1]. The sensor-clouds are specially used for military applications such as battlefield surveillance, industrial and consumer applications such as machine health monitoring, industrial control and process monitoring, etc. [2]. The main challenges of the sensor-clouds are limited power supply and the networks’ life span. A major problem of sensor-clouds is that non-chargeable batteries are used as an energy supply on each node, so once nodes are deployed it cannot be recharged or replaced. Many cluster-based routing algorithms are developed in sensor-clouds, mainly to minimize the energy consumption and trying to maximize the lifespan [3] of the whole network.

In the era of engineering, the decision making problems are solved by using MCDM [4] technique with multiple attributes. In this method multiple alternatives are compared and ranked based on the degree of their respective attributes. AHP is one of the MCDM techniques which was developed by Saaty (1977, 1994) [5]. AHP consists of a decision matrix, where is the number of alternatives and is the number of decision criteria [6].

The paper is organized as follows: Section 2 introduces the routing protocol. Section 3 discusses the survey. Section 4 discusses the proposed method. Section 5 analyzes the simulation results. Lastly, Section 6 concludes the work.

2. Routing protocols in sensor-clouds

Various types of routing protocols in sensor-clouds have been proposed by researchers. Three way routing protocols are classified on the basis of protocol operation, network structure and packet destination [7]. Figure 1. shows the classification of routing protocols in the sensor-cloud. The proposed work is based on this network structure.

Figure 1.

Classification of routing protocols [8].

The hierarchical-based routing [9] protocols, the whole networks are divided into several smaller regions. Each small region is called a cluster. Every cluster has a cluster head (CH), which acts as a managing node and the rest of the nodes are member nodes. CH acts as a communication bridge between BS and member nodes. The merit of this technique is the gathering of data. CHs are collecting the data from all member nodes and sent collected data to the BS [10]. The major drawback of this routing model is that energy consumption of the CHs are increased due to their additional functions. Mostly, CH formation depends on a single criterion. This criterion is that CH is generally elected on the basis of randomly or residual energy, sink distance or node density. If the head node is chosen on the basis of residual energy, but the higher energy node is too far away to sink, then the problem appears for election on CH. Similarly, the same problem will be arise when the distance is shorter to the sink and CH [23] is created near the BS. So, single criterion is not enough for head node chosen. Multiple criterion is then proposed to overcome this problem. The energy saving method is discussed below.

2.1 Network structure routing protocol division

Network structure acts as a significant role in sensor-clouds routing protocols. It can be divided in three ways: location-based, hierarchical-based/cluster-based, and flat-based.

In flat-based routing all nodes have the same role. Sink sends queries to nodes and nodes are transmitting data to BS. In location-based protocol nodes are calculating the distance from location information for the next hop selection. In the hierarchical-based/cluster-based routing [9] protocols, the whole networks are divided into several smaller regions. Each small region is called a cluster. Every cluster has a CH, which acts as a managing node and the rest of the nodes are member nodes. CH acts as a communication bridge between BS and member nodes. The merit of this technique is the gathering of data. CHs are collecting the data from all member nodes and send collected data to the BS [10]. The energy saving method is discussed below.

2.2 Energy saving method in sensor-cloud clustering

The energy saving approach in the sensor-cloud is very important because forming clusters and selecting CH efficiently preserve energy. These two are important constraints on the clustering mechanism in the sensor-cloud. In the three step cluster formation rotation, CH selection and rotation and intra-cluster communication energy efficient clustering happened.

The network life span and stability depends on the clustering and head selection. The popular routing protocols that are available are LEACH [11], TEEN [12], SEP [13], DEEC [14].

Each protocol maintains their CHs (Table 1).

Table 1
Comparison of sensor-cloud hierarchical-based protocol

Protocols	Energy	Scalability	Energy	Network	Data	CH selection
	homo/hetero- genetic		efficiency	lifetime	delivery model	Initial energy	Residual energy	Average energy of network
LEACH	Yes	Very limited	Lower	Poor	CH	Yes	No	No
TEEN	Yes	Very limited	Very high	Best	Threshold value	Yes	No	No
SEP	No (Heterogeneous in two level)	More	Low	Good	CH	No	Yes	No
DEEC	No (Heterogeneous in multi level)	Most	High	Better	CH	No	Yes	Yes

3. Literature survey

If we want to improve network life span in the sensor-cloud, we have to consider appropriate clustering methods for CH election. Different kinds of clustering algorithms are developed for this purpose.

In 1981 Hwang and Yoon developed first single-hop clustering algorithm LEACH. In this algorithm entire networks are divided into several clusters. In every cluster one node act as a head node and the other nodes act as member nodes. CHs are consuming more energy than member nodes. If a node continuously acts as CH, it can die quickly. This problem is solved by the LEACH algorithm, in which the nodes are dynamically changed. The LEACH protocol works in rounds. One round has two phases, one is a setup phase and the other is a steady state phase. In the setup phase clustering and CH formation happen. The communication between CH and BS occurs in the steady state phase [15].

In 2002 Heinzelman et al. improved the traditional LEACH algorithm and developed the C-LEACH or ‘centralized LEACH’ protocol. Here, sink nodes are elected as the CHs. It gives better results than LEACH [16].

In 2004 Younis and Fahamy developed the Hybrid Energy-Efficient Distributed (HEED) algorithm. On the basis of the remaining energy of nodes CHs are chosen [17].

In 2006 Kim and Chung proposed an algorithm which focusses on LECH-M or ‘LEACH-Mobile’. Here, networks and nodes are mobile. Nodes are communicated through CH [18].

In 2013 Azada proposed a technique called fuzzy multiple attribute decision making (MADM) approach of CH election. This includes two criteria that are used for selecting CH election through selection and they observed that every round head is changing. Here, LEACH-M are considered [19].

In 2017 Khan and Young proposed fuzzy-TOPSYS-based CH election in mobile SNs. Here, four criteria are chosen. CHs are elected and compared with the traditional LEACH and conventional fuzzy [20].

In 2018 Chen et al. proposed clustering algorithm on the basis of K-means and PROMETHEE methods. Here, ordered K-means algorithm is discussed and solves the human development indexing problem [21].

The different methods of energy efficient protocol are discussed above. In our proposed method we define a threshold value for change of CH, so, CHs are not changing in every round and for this reason control overhead is reduced. Here, we consider four criteria for election of CH in the multi-hop model. So that the proposed method gives energy efficient, long life, less CH formation and control overhead approach in intra-inter-clustering communication.

4. Proposed clustering scheme based on AHP

Our scheme is based on the popular LEACH protocol and it involves four phases: the network deployment phase, neighbor discovery phase, cluster formation and CH election phase, and communication phase. Each of these phases are discussed below in Subsections 4.1–4.4.

4.1 Network deployment

Network deployment is the first step of the proposed scheme, which is shown in Fig. 2. The consideration nodes are homogeneous random in nature and fixed uniformly deployed. It is also assumed that the dimension of the field and BS coordinates are also known to us. The BS is easily handled by aggregating, transmission and reception data from CHs to the destination.

Figure 2.

Network deployment.

4.2 Neighbor discovery

The second step of the proposed scheme is the neighbor discovery phase. Initially, a ‘Hello’ packet will be broadcast by the nodes and every node contains their node ID: C1 residual energy, C2 node density, C3 distance to the sink or BS, average distance between nodes and its neighbors C4. Initially, ‘Hello’ packet is empty in C2 and C4 field because the node does not have the neighbor information. After node ID sharing C2 and C4 exchange their information. All nodes $T_{r}$ transmission range receive the packet and the nodes are updated, their neighborhood table ( $T$ ). Assume, there are neighbors of node $k$ , then $T_{k}$ will be a $(n+1)\times 4$ matrix:

${T_{k}}=\begin{array}[]{@{}c@{}}{{a_{1}}}\\ {{a_{2}}}\\ {{a_{3}}}\\ {\vdots}\\ {{a_{4}}}\end{array}\begin{pmatrix}{{C_{1}}}&{{C_{2}}}&{{C_{3}}}&{{C_{4}}}\\ {{v_{1,1}}}&{{v_{1,2}}}&{{v_{1,3}}}&{{v_{1,4}}}\\ {{v_{2,1}}}&{{v_{2,2}}}&{{v_{2,3}}}&{{v_{2,4}}}\\ {{v_{3,1}}}&{{v_{3,2}}}&{{v_{3,3}}}&{{v_{3,4}}}\\ {\vdots}&{\vdots}&{\vdots}&{\vdots}\\ {{v_{n+1,1}}}&{{v_{n+1,2}}}&{{v_{n+1,3}}}&{{v_{n+1,3}}}\end{pmatrix}$

After updating, the MCDM technique is used to evaluate the rank, index and distribute it to all neighbor nodes through the ‘Hello’ packet. The rank index will be calculated using AHP.

4.3 CH selection and cluster formation

Comprehensive explanation of the CH selection process of our proposed scheme is given in this section. The following steps are performed to calculate rank index using AHP [22]:

Step 1: In the AHP is the estimation of pertinent data. The estimation of the $C_{ij}$ and $W_{k}$ values of the decision matrix. The decision maker has to eliminate from consideration that score 10 percent of another alternative. Table 2 shows the scale of relative importance according to Saaty (1980).

Table 2
Scale of relative importance

Intensity of weight values	Definition	Explanation
1	Same situation for CH	Two criteria contributed equally to formation CH
3	Weakly suitable for CH	Criteria slightly suitable one node over another node
5	Strongly suitable for CH	Criteria strongly suitable one node over node
7	Conform for CH	A criteria are giving evidence with good cooperation in the election of CH
9	Not suitable for CH	The criteria favoring one node over node are of the highest possible order of affirmation
2, 4, 6, 8	Intermediate values between the two criteria	When CH chosen comparison is needed
Reciprocals of above nonzero	If activity $i$ has one of the above nonzero numbers assigned to it when compared with activity $j$ , then $j$ has the reciprocal value when compared with $i$

Pairwise comparison, the relative importance of one criterion over another can be expressed.

${T_{k}}=\begin{pmatrix}1&{1/2}&{1/6}&{1/7}\\ 2&1&{1/3}&{1/5}\\ 6&3&1&{1/4}\\ 7&5&4&1\\ \end{pmatrix}$

Step 2: On the basis of Saaty scale, calculate the weighted matrix $W_{k}$ . The matrix is shown below:

${W_{k}}=\begin{pmatrix}{0.0558928}\\ {0.09948}\\ {0.0248192}\\ {0.596435}\end{pmatrix}$

Step 3: Calculate the Consistency Indexed (CI). Formula of the CI indexed is,

$CI=\frac{{\lambda_{\max}}-n}{n-1}$

Here, CI $=$ 0.0514.

Step 4: Finally, calculate the Consistency Ratio (CR). According to Saaty it is less than or equal to 0.1. The CI of a randomly generated pairwise comparison matrix, where ‘n’ is the order of matrix and random inconsistency indices for ‘n’ $=$ 10 (Saaty, 1980).

$n$	1	2	3	4	5	6	7	8	9	10
RI	0.00	0.00	0.58	0.9	1.12	1.24	1.32	1.41	1.46	1.49

Then formula of CR is shown in equation

$CR={CI}/{RI}$

Here, $CR=$ 0.0571.

Example. Assume that we consider four nodes in a cluster. Nodes are denoted as $m_{1}$ , $m_{2}$ , $m_{3}$ and $m_{4}$ . Now we calculate pairwise comparison matrix, i.e., ${\bm{T}}_{{R}}$ , ${\bm{T}}_{{N}}$ , ${\bm{T}}_{{D}}$ and ${\bm{T}}_{{A}}$ , in terms of residual energy, nodes density, distance to the sink and average distance between nodes and its neighbors of each alternative nodes over other nodes.

$\displaystyle{T_{R}}=\begin{array}[]{c}{{m_{1}}}\\ {{m_{2}}}\\ {{m_{3}}}\\ {{m_{4}}}\end{array}\begin{pmatrix}{{m_{1}}}&{{m_{2}}}&{{m_{3}}}&{{m_{4}}}\\ {1/1}&{1/4}&{4/1}&{1/6}\\ {4/1}&{1/1}&{4/1}&{1/4}\\ {1/4}&{1/4}&{1/1}&{1/5}\\ {6/1}&{4/1}&{5/1}&{1/1}\\ \end{pmatrix}$ $\displaystyle{T_{N}}=\begin{array}[]{c}{{m_{1}}}\\ {{m_{2}}}\\ {{m_{3}}}\\ {{m_{4}}}\end{array}\begin{pmatrix}{{m_{1}}}&{{m_{2}}}&{{m_{3}}}&{{m_{4}}}\\ {1/1}&{2/1}&{5/1}&{1/1}\\ {1/2}&{1/1}&{3/1}&{2/1}\\ {1/5}&{1/3}&{1/1}&{1/4}\\ {1/1}&{1/2}&{4/1}&{1/1}\\ \end{pmatrix}$ $\displaystyle{T_{D}}=\begin{array}[]{c}{{m_{1}}}\\ {{m_{2}}}\\ {{m_{3}}}\\ {{m_{4}}}\end{array}\begin{pmatrix}{{m_{1}}}&{{m_{2}}}&{{m_{3}}}&{{m_{4}}}\\ {1/1}&{3/1}&{1/3}&{1/3}\\ {1/3}&{1/1}&{1/1}&{1/3}\\ {3/1}&{3/1}&{1/1}&{3/1}\\ {3/1}&{3/1}&{3/1}&{1/1}\\ \end{pmatrix}$ $\displaystyle{T_{A}}=\begin{array}[]{c}{{m_{1}}}\\ {{m_{2}}}\\ {{m_{3}}}\\ {{m_{4}}}\end{array}\begin{pmatrix}{{m_{1}}}&{{m_{2}}}&{{m_{3}}}&{{m_{4}}}\\ {1/1}&{1/3}&{3/1}&{1/3}\\ {3/1}&{1/1}&{1/3}&{3/1}\\ {1/3}&{3/1}&{1/1}&{3/1}\\ {3/1}&{1/3}&{1/3}&{1/1}\\ \end{pmatrix}$

Then the vector determines the rank of the matrix from the comparison matrix, i.e., $\bm{W}_{R}$ , $\bm{W}_{N}$ , $\bm{W}_{D}$ and $\bm{W}_{A}$ .

$\displaystyle{W_{R}}=\begin{pmatrix}{0.1160}\\ {0.2470}\\ {0.0600}\\ {0.5770}\end{pmatrix}\quad{W_{N}}=\begin{pmatrix}{0.3790}\\ {0.2900}\\ {0.0740}\\ {0.2570}\end{pmatrix}$ $\displaystyle{W_{D}}=\begin{pmatrix}{0.1720}\\ {0.2240}\\ {0.3280}\\ {0.2760}\end{pmatrix}\quad{W_{A}}=\begin{pmatrix}{0.2400}\\ {0.2760}\\ {0.3280}\\ {0.1720}\end{pmatrix}$

Then calculate the highest ranked of the node, which act as a CH

$\begin{array}[]{c}{{m_{1}}}\\ {{m_{2}}}\\ {{m_{3}}}\\ {{m_{4}}}\end{array}\begin{pmatrix}{{C_{1}}}&{{C_{2}}}&{{C_{3}}}&{{C_{4}}}\\ {{W_{1,R}}}&{{W_{1,N}}}&{{W_{1,D}}}&{{W_{1,A}}}\\ {{W_{2,R}}}&{{W_{2,N}}}&{{W_{2,D}}}&{{W_{2,A}}}\\ {{W_{3,R}}}&{{W_{3,N}}}&{{W_{3,D}}}&{{W_{3,A}}}\\ {{W_{4,R}}}&{{W_{4,N}}}&{{W_{4,N}}}&{{W_{4,A}}}\end{pmatrix}*\begin{pmatrix}{{% C_{R}}}\\ {{W_{1,K}}}\\ {{W_{2,K}}}\\ {{W_{3,K}}}\\ {{W_{4,K}}}\end{pmatrix}\begin{array}[]{c}{{C_{1}}}\\ {{C_{2}}}\\ {{C_{3}}}\\ {{C_{4}}}\end{array}$

$\begin{array}[]{c}{{m_{1}}}\\ {{m_{2}}}\\ {{m_{3}}}\\ {{m_{4}}}\end{array}\begin{pmatrix}{0.1916}\\ {0.2629}\\ {0.2145}\\ {0.1673}\end{pmatrix}$

So that,

${C_{2}}>{C_{1}};∼{}∼{}{C_{2}}>{C_{3}};∼{}∼{}{C_{4}}>{C_{3}};∼{}∼{}{C_{4}}>{C_{% 1}}$

And,

${m_{2}}>{m_{3}}>{m_{1}}>{m_{4}}$

The average distance between nodes and its neighbors is the highest ranked criteria of CH forming and node 2, i.e. $m_{2}$ , is highest ranked to make a CH.

The node with the highest value in this CI act as a CH in that region. In this region other nodes are member nodes. When the join requests are successfully received, CH allows all member nodes. In this process clusters are made in the whole network.

Figure 3.

CH change procedure [8].

Once a round of clusters are formed, member nodes are communicating with the CHs. If the index value of CH is less than the index value of any nodes addition of the threshold value (here 0.1), then CHs could not act as CH nodes, then further election processes will be started. The flowchart (Fig. 3) shows the CH change procedure. In this way we can control the frequent changing of heads in every round. This process still continues until all nodes have died in the particular network.

4.4 Communication phase

The communication phase will start after the election of CH and creation of the cluster. In reality, we considered the multi-hoping communication model in our work. We considered that CH receives the data directly from the nodes in the five meter range, greater than this range multi-hopping is used for CH communication. Same criterion is followed for CH to sink communication. In this case the CH sink distance is twenty meters, greater than this range multi-hopping is used. In this scenario, multi-hopping occurs between the CH previous condition it will happen between the nodes. The multi-hop technique increases the network lifespan and stability. Figure 4. shows inter and intra- multi-hop-cluster communication.

Figure 4.

Intra and inter-multi-hop-cluster communication.

The complete flowchart of the proposed methodology is shown in Fig. 5.

5. Simulations

In simulations, comparisons will be done between LEACH and the proposed work. For simulation purposes, we have used MATLAB software. Here, we discuss simulation parameter and results.

5.1 Simulation environment

For simulation, the parameters shown in Table 3 are considered. Here, 100 nodes are assumed for the clustering purpose. Area of the network is (100 m $\times$ 100 m). In the beginning nodes are consuming 0.5 Joules. Target of the sink position is (50 $\times$ 100).

5.2 Results

Table 3
Simulation environment

Parameter	Value
Area of network	100 m $\times$ 100 m
n (number of nodes)	100
sn (sink position)	50 $\times$ 100
Packet size (data)	4000 bits
Packet size (Hello)	200 bits
Sink position	(1.5 m $\times$ 0.5 m $\times$ 0.5 m)
Initial energy	0.5 Joules
Data aggregation energy	50 pJ/bit
$E_{elect}T_{x}$ (transmitter electronics)	50 nJ/bit
$E_{elec}R_{x}$ (transmitter electronics)	50 nJ/bit
$E_{amp}$ (transmit amplifier)	100 pJ/bit/m ${}^{2}$

Figure 5.

Flowchart of proposed methodology.

We analyzed energy consumption and stability and compared the results with the existing LEACH protocol and proposed method by considering the above parameter. The results are discussed in from Fig. 6 to 10.

5.2.1 Network stability

Stability of the whole network is shown in Fig. 6. Approximately 170 rounds, the first node dies in LEACH but in the proposed work it dies around the 1600 round. Hence, the proposed technique is better than the traditional protocol on the basis of stability. Here, LEACH CH election process is based on one criteria, but in the proposed work CH election is based on every node decision.

Figure 6.

Network stability.

5.2.2 Network life span

The life span of the entire network is shown in Fig. 7. Approximately 1000 rounds, the last node dies in LEACH but in the proposed mechanism it occurs at 2400 rounds. It happens because of CH election, in our work it is based on multi criteria. Hence, our work is better than traditional protocol.

Figure 7.

Network life span.

5.2.3 Network energy consumption

Consumption of energy per round of the whole network is observed in Fig. 8. The proposed method consumes less energy than LEACH. The main reason of this is because for changing CH we consider a threshold value and for clustering communication (intra and inter) we use multi-hop model. Hence, LEACH is less energy efficient than the proposed work.

Figure 8.

Network energy consumption.

5.2.4 Network overhead

Figure 9 shows the control overhead of network comparison. It shows that the overhead of the control signal is smaller in our proposed method than the traditional LEACH. The main reason is minimization of overhead is distributed algorithm. It is used for rare changing of CH election.

Figure 9.

Network overhead.

Figure 10.

Throughput network.

5.2.5 Throughput of network

Figure 10 shows the throughput of network. From the figure we conclude that throughput is high in the proposed algorithm. The reason behind this is that the CH election and multi-hop-cluster model are used in the proposed work.

6. Conclusion

The proposed work described a new clustering technique based on multi criteria decision making using AHP. Here, we considered four criteria to form a CH. In this paper we chose to distribute algorithm, it means modes are taking the verdict to act as a CH or not. In the proposed method we can control the frequent changing of heads. If the index value of CH is less than the index value of any nodes addition of the threshold value (here 0.1) then CHs could not act as CH nodes, then further election processes will happe. This process still continues until all nodes have died in the particular network. The proposed method also improves the inter-and intra-multi-hop-cluster communication. Shorter range nodes directly send their data to the CH but long distance multi-hopping is required to communicate with CH. The same problem will happen in data transmission between CH to BS. In this way multi-hopping increases the network stability and life span. Using MATLAB we compared the existing LEACH with our proposed work and concluded that our work is better than the traditional work. This improved algorithm is achieved using the distributed MCDM algorithm and also using inter-and intra-cluster multi-hop communication model.

References

Chaurasiya

S.K.

Pal

and Bit

S.D.

, An enhanced energy-efficient protocol with static clustering for WS, International Conference on Information Networking (ICOIN), Kuala Lumpur, Malaysia, 2011, pp. 58–63.

Shah

Javaid

and Qureshi

T.N.

, Energy efcient sleep awake aware (EESAA), Intelligent Sensor Network Routing Protocol, arXiv preprintarXiv:1212.4106, 2012.

Heinzelman

Chandrakasan

and Balakrishnan

, Energy-efcient rout-ing protocols for wireless microsensor networks, in: Proc 33rd Hawaii Int Conf SystemSciences (HICSS), Maui, HI, 2000.

Seram

and Tzeng

G.-H.

, Compromise solution by MCDM methods: A comparative analysis of VIKOR and TOPSIS, European Journal of Operational Research (2004), 445–455.

Belton

and Gear

, On a short-coming of Saaty’s method of analytic hierarchies, Omega (1983), 228–230.

Boucher

T.O.

and McStravic

E.L.

, Multi-attribute evaluation within a present value framework and its relation to the analytic hierarchy process, The Engineering Economist 37 (1991), 55–71.

Al-Karaki

J.N.

and Kamal

A.E.

, Routing techniques in wireless sensor networks: A survey, IEEE Wireless Communications 11(6) (2004), 6–28.

Mustafa

and Criteria

, Decision making based clustering technique for WSNs, Thesis COMSATS Institute of Information Technology, (2013), 2–68.

Ateya

A.A.

Sayed

M.S.

and Abdalla

M.I.

, Multilevel hierarchical clustering protocol for wireless sensor network, International Conference on Engineering and Technology (ICET), 2014.

10.

Singh

S.P.

and Sharma

S.C.

, A survey on cluster based routing protocols in wireless sensor networks, International Conference on Advanced Computing Technologies and Applications 45 2015, pp. 687–695.

11.

Huang

Ling

and Zhou

, An improved LEACH routing algorithm for wireless sensor network, International Journal of Wireless Information Networks, (2018), 1–9.

12.

Yadav

J.S.

and Yadav

R.S.

, A review on energy efficient protocols in wireless sensor networks, Wireless Network 22 (2016), 335–350.

13.

Kumar

Verma

S.K.

and Kumar

, Enhanced threshold sensitive stable election protocol for heterogeneous wireless sensor network, Wireless Personal Communication 85(4) (2015).

14.

Singh

and Kumar

, Energy efficient heterogeneous DEEC protocol for enhancing lifetime in WSNs, Engineering Science and Technology, an International Journal 20 (2017), 345–353.

15.

Hwang

C.L.

and Yoon

, Multiple attribute decision making lecture notes in economics and mathematical systems, Springer-Verlag, Berlin 186 (1981), 1981.

16.

Heizelman

W.R.

Chandrakasan

and Balakrishnan

, An application-specific protocol architecture for wireless micro-sensor network, IEEE Trans Wireless Commun 1 (2002), 660–670.

17.

Younis

and Fahamy

, HEED: A hybrid, energy-effSaatyicient distributed clustering approach for adhoc sensor network, IEEE Trans Mobile Comput (2004), 660–670.

18.

Kim

D.S.

and Chung

Y.J.

, Self-organization routing protocol supporting mobile nodes for wireless sensor network, First International Multi-Symposium on Computer and Computational Sciences (IMSCCS06), 2006.

19.

Azada

, Cluster head selection in wireless sensor networks under fuzzy environment, ISRN Sens Netw 2013, Hindawi Publishing Corporation, Article ID 909086, 2013.

20.

Khan

B.M.

and Young

, Fuzzy-TOPSIS based cluster head selection in mobile wireless sensor networks, Journal of Electrical Systems and Information Technology, (2017).

21.

Chen

Wang

and Liu

, An ordered clustering algorithm based on K-means and the PROMETHEE method, International Journal of Machine Learning and Cybernetics 9(6) (2018), 917–926.

22.

Liu

and Qi

, Analytic hierarchy process based judgment model of highly cited papers in journal, IEEE 2nd International Conference on Big Data Analysis (ICBDA), 2017. DOI: 10.1109/ICBDA.2017.8078716.

23.

Rathi

and Acharjya

D.P.

, A framework for prediction using rough set and real coded genetic algorithm, Arabian Journal for Science and Engineering 43(8) (2018), 4215–4227.