Wireless Communications and Mobile Computing

Wireless Communications and Mobile Computing / 2019 / Article

Research Article | Open Access

Volume 2019 |Article ID 6981657 | https://doi.org/10.1155/2019/6981657

Chen Zhixiong, Cui Juanping, Yuan Jinsha, Xiao Nan, Han Dongsheng, "Network Access Selection Algorithm Based on Balanced Profits between Users and Network", Wireless Communications and Mobile Computing, vol. 2019, Article ID 6981657, 12 pages, 2019. https://doi.org/10.1155/2019/6981657

Network Access Selection Algorithm Based on Balanced Profits between Users and Network

Academic Editor: Mauro Femminella
Received02 Jan 2019
Revised16 Apr 2019
Accepted18 Apr 2019
Published02 May 2019


A network access selection algorithm based on the intuitionistic fuzzy analytic hierarchy process (AHP) and bilateral profit drive is proposed in this study for addressing problems regarding user–network bilateral profits. User preference, business demands, and network parameter changes are comprehensively considered in the algorithm. First, the initial weights centered at users are gained by intuitionistic fuzzy AHP. Second, the network participates in network access selection as a subject with competitive consciousness, and the entire selection process is transformed into a multiobjective optimization problem by the construction of a competitive model, thereby obtaining dynamic competitive weights. Third, the initial weights centered at users and the dynamic competitive weights are combined to obtain comprehensive weights. In this way, the dynamic adjustment of comprehensive weights is realized. Finally, candidate networks are ordered according to a comprehensive performance evaluation, and the optimal one is selected. The proposed algorithm is validated by simulation results to be valid in reducing the blocking rate of networks and optimizing network resource allocation. Therefore, it not only protects user–network bilateral profits but also maximizes comprehensive profits.

1. Introduction

With rapid development of wireless communication technology, users’ demands for communication businesses have shifted to diversity, personalization, and broad bands, which require the integration and collaboration of many wireless networks in providing users high-quality personalized services at any time and place [1, 2]. Increasing the utilization of network resources and providing users the best network access selection and the best service quality in heterogeneous wireless network environment have thus become important research topics [3, 4].

A network access selection algorithm based on the prediction of received signal strength was proposed in consideration of nonreal-time/real-time businesses covered by many networks [5]. However, this algorithm considers a single factor and thereby fails to provide users the most satisfying services. Thus, misjudgments can be made easily. Hence, the concept of context perception and the theory of multiattribute decision-making (MADM) were combined for ensuring high service quality to users [6]. At the same time, the weights of indexes at different levels are obtained through TOPSIS, and the best network is determined. MADM theory ensures that the algorithm generally meets the hierarchical demands of users. This theory has been extensively used in network access selection. A network access selection algorithm based on a utility function was proposed [7]. This algorithm quantifies the performance of candidate networks by using the improved TOPSIS algorithm and calculated user satisfaction to candidate network by utilizing the utility function, thereby ranking the candidate network. This algorithm could effectively reduce handover failures. The network access selection problem was converted into a constraint optimization one [8]. The best weight distribution mode of the evaluation indexes is found on the basis of the chaos genetic algorithm, thereby obtaining the optimal network. Under a simplified system architecture, [9] designed an operator-driven network selection scheme where an enhanced functionality of a mobile terminal is utilized, supporting a collaboration between the broker entity and devices. The scheme enables better use of information and better realization of network control and self-configuration. In [10] an intelligent selection of wireless networks was proposed for multimedia communication. This paper models the communication process as a multicriteria decision-making problem and uses the Analytic Network Process (ANP) to select the best network connection for the user. The aforementioned network access selection algorithms all use candidate networks as their objective evaluation objects without considering actual problems, such as individual profits at the network end and competitive game among networks.

However, decision-making based on the advantages of networks to users might cause cut-throat competition among networks and produce the ping-pong effect [11] because networks often exaggerate their advantages or conceal their disadvantages. The ping-pong effect may consequently influence the actual profit of users. Therefore, a network access selection algorithm based on intuitionistic fuzzy analytic hierarchy process (AHP) and bilateral profit drive (IFBP) is proposed for realizing balanced profits between users and networks. This study provides the following major contributions.

(1) In allusion to network access selection algorithms of MADM, a calculation method of comprehensive weights based on balanced user–network profits is presented with consideration of the weights of expert evaluation at the user end and of competitive weights at the network end.

(2) The intuitionistic fuzzy AHP is applied for calculating intuitionistic fuzzy numbers and hesitancy of index weights at the user end to reflect the subjectivity and uncertainty of expert evaluation. The fuzziness of weights reveals the degree of competition among different networks and could be used to determine the friction coefficient of networks.

(3) The network is considered a profit end of access selection and an individual with independent consciousness. The calculation of index weights at the network end is then converted into a mathematical problem of multiobjective optimization on the basis of the friction coefficient and a network competition model. Competitive weights ensuring the maximum profits of the network could be acquired by the solution of this mathematical model.

2. IFBP Algorithm

The proposed IFBP algorithm is a network access selection algorithm that is based on MADM. This algorithm is mainly composed of the construction of the index system, normalization of index parameters, and calculation of index weights. The primary goal is to construct an index evaluation system.

2.1. System Model and Index Evaluation System

The selection of decision-making indexes in the index evaluation system is vital to network access selection based on comprehensive performance evaluation. A total of 3–10 decision-making indexes must be selected [6]. Load, coverage area, delay, jitter, packet loss rate, safety, and cost are the major factors that influence network performance.

To protect the accuracy and complexity of the access selection algorithm, user’s concern, network performance, cost and operability, without loss of generality, are considered in this paper. In this study, six typical decision indexes, namely, received signal strength, service expense, network delay, network load, frame error ratio, and data transmission rate, were selected. A comprehensive evaluation index system was constructed (Figure 1). Investigation of the weight calculation method was emphasized, and evaluation indexes were adjusted according to practical situations.

In Figure 1, wij is the weight of the index j of the candidate network i, represents the initial value of the parameters of index j. i = 1,2,...,n (n is a positive integer that denotes the number of candidate networks), and j = 1,2,...,m (m=6 and denotes the number of evaluation indexes). The initial values of index parameters have to be normalized for comparability in the evaluation. Indexes can be divided into profit and cost types according to attributes, which shall be normalized by different methods.

The profit type of indexes can be expressed as

(2) The cost type of indexes can be expressed aswhere is the initial value of the parameters of the evaluation index j in the candidate network i before normalization. is the index value after normalization and forms the normalization matrix .

The key step of network access selection algorithm based on MADM is calculating the comprehensive performance value of the network i according to wij and xij.

Finally, the network with the maximum performance value is selected from n candidate networks.

2.2. Calculation of Fuzzy Constant Weight at the User End Based on Intuitionistic Fuzzy AHP

The proposed IFBP algorithm considers the network an intelligent agent that has an independent consciousness to protect the user–network bilateral profits. This algorithm calculates the constant weight at the user end and the dynamic competitive weight at the network end in accordance with different networks and state indexes. Afterward, the comprehensive weights of the network indexes are calculated by weight factor and equilibrium weighting.

The traditional calculation methods of constant weight mainly include Entropy weight method, AHP, and the structural entropy weight method. The characteristics of these methods are summarized in Table 1.

Weight calculation methodsAdvantages and disadvantagesReferences

Grey relational analysisIt can avoid wrong weight distribution caused by conflicting opinions of experts, but determining the identification coefficient is difficult.[1113]

Entropy weight methodIt can significantly decrease the influences of experts’ subjectivity on quantification of weights, but reasonably explaining the quantification results of weights is difficult.[7, 14]

AHPAHP can be used in the quantitative analysis of qualitative indexes and achieves clear logics. However, it shows strong subjectivity and needs to test the consistency of the judgment matrixes. It is also inapplicable to multilayer index systems.[6, 11]

Analytic network AHPIt possesses the advantages of AHP and considers the relationship of indexes but requires computer processing data and is inapplicable to multilayer index systems.[11, 13, 15]

Structure entropy weight methodThe structure entropy weight method can not only relieve the influences of subjective preference on weight distribution but also protect the objectivity of weights and is easily calculated. Nevertheless, the random value of conversion parameters may bring different degrees of “conversion distortion” and affect weight distribution.[16, 17]

Intuitionistic fuzzy AHP was applied to calculate the weights of indexes to reflect the subjectivity and uncertainty of expert evaluation [18]. This method reflects the subjective judgment of decision-makers on research objects and makes quantitative representation of abstention or hesitancy of experts. Therefore, it is flexible and practical in processing these uncertainties. The calculation of constant weight based on the intuitionistic fuzzy AHP covers the three following steps.

Step 1 (intuitionistic fuzzy judgment matrixes are constructed). Experts conduct pairwise comparison of indexes on the basis of the comprehensive evaluation index system in Figure 1 and finally obtain the intuitionistic fuzzy judgment matrix A.where aij = (uij, vij) (i, j=1,2,...,m). During the comparison of the importance of indexes i and j, uij represents the decision-maker preferences to i, and vij denotes the decision-maker preferences to j. reflects the hesitancy or uncertainty of decision-makers. uij, , and uij + vij ≤ 1.

Step 2 (a consistency intuitionistic fuzzy judgment matrix is constructed on the basis of matrix, and a consistency test is performed subsequently). (1) The calculation of matrix elements is divided into three situations.
(a) When j > i +1, (b) When or , .
(c) When .
(2) Consistency test: the distance function (d) (between A and ) and the threshold () of the consistency index are defined. When , matrix A meets the consistency condition. When A fails to satisfy the consistency condition, it could be adjusted by the similarity factor such that it fulfills the consistency condition.

Step 3 (the initial constant weight vectors of indexes are determined). The index weight vectors could be obtained when A meets the consistency condition. is the intuitionistic fuzzy number that corresponds to the weight of index i and could be expressed as . and could be expressed aswhere denotes the importance of index i relative to other indexes. reflects the nonimportance of index i and meets , , and .

2.3. Dynamic Network Access Selection Algorithm Based on Bilateral Profit Drive

Networks are used as “intelligent agents” with competitive consciousness; their advantages are developed thoroughly for gaining the top rank in the competition and attracting as many users as possible to protect their fair competition and the full expression of reasonable demands of different networks. These approaches are conducive to protect their profits. Thus, each network has to highlight their advantages and weaken the advantages of others.

In this study, the competition of networks is divided into two subprocesses to calculate the competitive weights (this process is centered at subnetwork and solved by multiple models in parallel). For example, the competitive weight of network i is used as the solving target in mathematical modeling. The network is divided into network i and the rest network. The goal of the competitive weight is to maximize the performance score of its own network and minimize the performance score of the other networks. The competitive weight of networks is dynamic because the values of actual index parameters in different networks are dynamic, thereby realizing the dynamic updating of the comprehensive weight.

The specific steps of the dynamic network access selection algorithm that is based on bilateral profits are as follows.

Step 1 (construct a competition model). Each network shall formulate a group of index weights according to their advantages and the characteristics of others at the current moment to access many users. This group of weights changes with network parameters. Thus, it can improve its advantages and weaken those of potential competitors.

The normalized matrix of all network indexes is . Then, the actual appeals and state of the network shall be considered. A multiobjective optimization mathematical model can be constructed for network i to solve the competitive weight .where nk is the number of networks in the rest network set Nk and is the normalized value of index j of network l in Nk. The friction coefficient between networks i and l is defined as fil. The competition between networks i and l is strong when their final evaluation values are close. Thus, the friction coefficient is accordingly high.

Problem (8) is a mathematical model of multiobjective optimization, where can be calculated by (9)-(11) and and xij are values in the normalized matrix . Besides, there are only linear equations and linear inequalities in model (8). We can use the linear programming method to obtain the network-side competition weights . Thus, the complexity of problem (8) is acceptable and can be solved by a short time requested by modern interactive services. In the simulation, we can use the fgoalattain function in Matlab to solve the model with two seconds.

Step 2 (calculate friction coefficient with improvement on the basis of intuitionistic fuzzy AHP). Equation (7) shows that fuzzy numbers and of the index weights can be obtained from intuitionistic fuzzy AHP, which reflects the uncertainties of weight values. In other words, network performance has intersections attributed to index weight boundaries. The algorithm determines the weight boundaries by intuitionistic fuzzy numbers to calculate the friction coefficient of different networks.
First, hesitancy is calculated according to and . is the uncertainty with which experts judge the importance of the index i. The maximum and minimum weights of the index i are further determined by hesitancy . This phenomenon reveals that the weight of index i is
Second, the value range of the comprehensive performance of networks i and l can be expressed as follows according to the upper and lower boundaries of the index weights:Therefore, the friction strength (sil) between networks i and l isFor l ∈Nk. If , then network l is not a competitor of network i. Thus, = 0 and = 0. If , then can be calculated by

Step 3 (calculate the comprehensive weight). The proposed IFBP algorithm performs linear weighting on the initial constant weight of users and the multiobjective optimized competitive weights to gain the comprehensive weight.
The proposed algorithm calculates the scores of different index weights through the similarity function according to and in (7) because the initial constant weight is an intuitionistic fuzzy number. Finally, the determined numerical values of the initial weights of indexes are obtained from normalization.Subsequently, the comprehensive weight () is obtained from the linear weighting.where and (i, k =1,...,n; j=1,...,m) are weight factors, , . A weight factor matrix is composed of and meets . To ensure user–network bilateral profits and the reflection of reasonable competition among different networks without loss of generality, , , and for network i in the follow-up simulation analysis.

Step 4 (calculate the comprehensive evaluation matrix of networks, and determine the ranking). The comprehensive evaluation matrix Qi of different networks could be obtained from (14) on the basis of the normalized decision-making matrix in Section 2.1 obtained by (1) and (2) and the comprehensive weight vector matrix in this section.

Finally, the basic procedures of IFBP algorithm are shown in Figure 2.

(1) An evaluation index system centered as users is constructed in accordance with user preference and service demands.

(2) With consideration of decision uncertainties, the initial constant weights of indexes are calculated by intuitionistic fuzzy AHP. The fuzzy vectors of index weights are calculated according to the importance of pairwise indexes determined by the expert group.

(3) Later, the network is viewed as a competitive intelligent agent, and the friction coefficient of the network is defined by the fuzzy boundary of the initial constant weight. Next, a multiobjective optimized competitive model is constructed in accordance with the real-time network state. Finally, the competitive weights of different networks are calculated.

(4) The comprehensive weights of network indexes are calculated by the weight factor and balanced weighting in accordance with the initial constant and competitive weights of indexes.

(5) Comprehensive evaluation values of different networks are calculated and ranked according to the network state and comprehensive weights. The networks in front ranks are selected for access.

3. Simulation and Analysis of the Proposed IFBP Algorithm

3.1. Simulation Scenes

A heterogeneous network is formed by overlapping and covering different wireless accessed networks. In this study, heterogeneous wireless network integration scenes were constructed by LTE, WiMAX, and WiFi technologies for verifying the performance of the proposed IFBP algorithm. Users were randomly distributed in the coverage area of the networks. Simulation mainly focused on the overlapping region of the three networks, and the simulation scenes are shown in Figure 3.

The physical layers of WiMAX, LTE, and WiFi all employ the OFDM modulation technique. The total numbers of two-dimensional resource units in the three networks were 213, 106, and 50, respectively. If new services arrived at the Poisson’s distribution in the overlapping region, then the arrival rate would be λ=0.1–1, and the service time would meet the negative exponential distribution (mean u=10 s). The simulation time was set as 200 s. Table 2 shows the basic simulation parameters of the different networks [19]. The parameters of the different judgment indexes in the simulation were randomly produced in different intervals because the actual network parameters were dynamic. The received signal strength of the networks could be acquired according to the propagation loss model due to the different carrier frequencies and transmitted powers of LTE, WiMax, and WiFi [20].

Type of networksLTEWiMAXWiFi

Carrier frequency (Ghz)
Transmitting power
Coverage area
Data rate (Kbps)
Delay (ms)
Frame error ratio
Network resources21310650

3.2. Simulation Results and Discussion

Without loss of generality, the following intuitionistic fuzzy judgment matrix A was applied to indexes and their ranking in Figure 1 according to the nine-level importance scale of indexes throughout the algorithm.

The values in the intuitionistic fuzzy decision matrix  A is denoted and determined by decision-maker preferences, during the comparison of the importance of indexes between each other. For example, the element in the matrix A indicates a quantized value of importance comparison when the index of service expense is compared to the delay index. In (15), is quantized and set as (0.8, 0.15) according to the nine-level quantization table in the literature [18]. It means that the experts believe that in general, the users pay more attention to the service expense and consider the expense is more important than the delay. From the value of =(0.8, 0.15), there is a degree of hesitation of 0.05=1-0.8-0.15, indicating an uncertainty of the experts judgment. For different scenarios and network service, the importance degree and the elements in matrix A can be adjusted and determined by experts accordingly. For example, in the URLLC scenario of 5G, the delay index is more crucial than service expense index, and the value of would be quite different. Meanwhile, for the coming 5G services, the indexes evaluation system may also be adjusted in terms of different scenarios accordingly.

The entire algorithm process is mainly based on the change of weights. Figure 4 shows the radar chart of the weights of each network index and thus visually reflects the weight change of the whole process.

The comprehensive weight realizes the trade-off between the user-end initial weight and the network-end competitive weight. Taking the LTE network as an example, the initial weight was mainly based on user concerns, and the weights of tariff and data rate were relatively large. However, the index weights of the LTE network obtained through the competition model were mainly concentrated on the frame error rate because the LTE network had the lowest frame error ratio among the networks. To increase its comprehensive evaluation value, the LTE network set the weight of the frame error ratio to be large while setting the weight of the tariff to be very small, thus concealing the disadvantage of high cost. In view of the interests of users and of the network, the comprehensive weights tended to balance the weights of the network indicators.

MATLAB simulation and analysis of the performance of the AHP algorithm [6], mobility load balancing (MLB) algorithm [21], and the proposed IFB algorithm in the overlapping regions of LTE, WiMAX, and WiFi were conducted from the perspectives of blocking rate, local balancing, user–network profits, and comprehensive profits. Load balancing was expressed by the Jain fairness index [22]; user profit [19] was defined as the weighted average of four indexes (data rate, delay, frame error ratio, and expense) gained by each user; data rate was normalized by profit, and rest indexes were normalized by cost indexes; network profit was defined as the weighted average of three indexes (expenses, blocking rate, and data rate) after each user was accessed into the network; expense was normalized by profit indexes, whereas the blocking and data rates were normalized by cost indexes; the comprehensive profit was equal to the weighted average of user and network profits. In order to make a fair comparison of the related algorithms, every simulation under the same parameter is repeated for many times, and finally the simulation results are averaged and presented in the figures from Figure 5 to 8.

Figure 5 shows the relation curves between the blocking rate of the network and the arrival rate of the new services in the three algorithms. First, the blocking rate of all three algorithms significantly increased with the arrival rate of new services. Second, the AHP, IFBP, and MLB algorithms began to suffer network blocking at λ values of 0.4, 0.5, and 0.8, respectively. Third, the AHP algorithm achieved the highest blocking rate under the same arrival rate of new services, followed successively by IFBP and MLB.

This result might be explained as follows. Users in the AHP algorithm select the best network according to individual preference. Many users might opt for the same network, thereby resulting in an increased early blocking rate. MLB considers load balancing and always selects the network with small loads. It achieves the minimum blocking rate of networks at the cost of users’ concerns. The blocking rate of the proposed IFBP algorithm was between those of AHP and MLB because it considers the profits of users and networks.

The relation curves between the load balancing of the three algorithms and the arrival rates of new services are shown in Figure 6. Load balancing is expressed by the fairness index (FI), which ranges between 0 and 1. A high FI reflects a high load balance among different networks. In Figure 5, the load balancing of IFBP and MLB is positively related with the arrival rate of new services, indicating that these two algorithms are significantly superior to the AHP algorithm. The reason is that MLB applies the global optimization strategy, which targets the load balance of the entire network. Only networks with small loads are accessed to users. The IFBP algorithm uses networks as subjects. Networks compete and balance mutually during access selection. Users always access networks with good comprehensive performance and comprehensive consideration of user–network bilateral profits. The competition and balancing mechanism ensures that the proposed IFBP algorithm is superior to the AHP algorithm in terms of the load balance of networks. Although the load balance of the IFBP algorithm is poorer than that of the MLB algorithm, the proposed IFBP algorithm can still consider user profits, and the acquisition of load balance reflects that the network autonomy is closer to that in practical situations.

The relation curves of the average user–user bilateral profits of the three algorithms and the arrival rate of new services are presented in Figure 7.

User profits: (a) the proposed IFBP and AHP algorithms are better than the MLB algorithm in view of the average user profits because they consider user preference and can effectively protect user benefits. (b) Unlike the AHP algorithm, the IFBP algorithm not only considers user preference and service demands but also has to solve the competition and balance problems of different networks. Given adequate network resources, network balance may affect user profits. In Figure 6, the average user profit of the proposed IFBP algorithm is slightly lower than that of the AHP algorithm when . (c) With the further increase of the arrival rate of new services, the IFBP algorithm improves blocking rate and load balancing better than does the AHP algorithm due to the network competition mechanism. Therefore, the user profit of the IFBP algorithm decreases more slowly than that of the AHP algorithm. The user profit of the IFBP algorithm is higher than that of the AHP algorithm when .

Network profits: (a) the AHP algorithm shows the lowest network profit because it decides according to user preference and neglects the actual parameter changes of the network. (b) The MLB algorithm helps users access LTE and WiMAX with many available network resources, thus achieving the highest network profit on the basis of the load balance perspective. (c) Meanwhile, the IFBP algorithm has to consider the profits of users and networks comprehensively. Therefore, the network profit of IFBP is significantly higher than that of AHP. The difference of the network profits between IFBP and MLB increases gradually with λ after λ=0.5.

The variation curves of the comprehensive profits of three profits with the arrival rate of new services are shown in Figure 8. In order to illustrate the credibility of the algorithm performance, we also show the interval of confidence of the comprehensive profits under different service arrival rates with a confidence level 95%. Comprehensive profit is defined as the weighted average of the user and network profits. The average comprehensive profits of the three algorithms are negatively correlated with the arrival rate of new services. Therefore, the IFBP algorithm is superior to the two other algorithms in terms of comprehensive profits, and this advantage is attributed to its simultaneous consideration of user and network profits. Simultaneously, in the case of confidence level 95%, the comprehensive profit under different service arrival rates basically falls in the middle of the confidence interval in the figure. It indicates that the simulation results are credible at the current confidence interval level.

4. Conclusions

A network access selection algorithm based on IFBP is proposed in this study. First, the proposed IFBP algorithm calculates the initial weights of indexes by centering at users and then views networks as individuals with competitive consciousness, thereby obtaining the competitive weights. Second, the comprehensive weights of indexes are acquired by the weighing of a certain proportion of the initial and competitive weights. Finally, the comprehensive evaluation values of different networks are calculated according to the comprehensive weights, and the network with the highest comprehensive evaluation value is selected for access. The simulation results demonstrate that the proposed IFBP algorithm can reduce the blocking rate of networks and effectively balance loads among networks. It can maximize comprehensive profits by simultaneously ensuring the profits of users and networks.

In the coming scenarios of 5G, due to the application of MEC, NFV, and SDN technology, the network configuration is more flexible, the real-time and reliability performance of the network are improved, and users have more and better access options. However, as the business amount increases, the service cost and power consumption for user terminals will also increase. Under the condition of limited communication, storage, and computing resources of network side, the performance improvement of network access selection by means of big data, game theory, power optimization, and other methods is a possible research content in the future.

Data Availability

The data used to support the findings of this study are available from the corresponding author upon request.

Conflicts of Interest

The authors declare that they have no conflicts of interest.


The research is supported by the National Natural Science Foundation of China (Nos. 61601182 and 61771195), Natural Science Foundation of Hebei Province (F2017502059 and F2018502047), and Fundamental Research Funds for the Central Universities (No. 2019MS088).


  1. P. Naghavi, S. H. Rastegar, V. Shah-Mansouri, and H. Kebriaei, “Learning RAT selection game in 5G heterogeneous networks,” IEEE Wireless Communications Letters, vol. 5, no. 1, pp. 52–55, 2016. View at: Publisher Site | Google Scholar
  2. Y. He, M. Chen, B. Ge, and M. Guizani, “On WiFi offloading in heterogeneous networks: various incentives and trade-off strategies,” IEEE Communications Surveys & Tutorials, vol. 18, no. 4, pp. 2345–2385, 2016. View at: Google Scholar
  3. Y. Qi, H. Wang, L. Zhang et al., “Optimal access mode selection and resource allocation for cellular-VANET heterogeneous networks,” IET Communications, vol. 11, no. 13, pp. 2012–2019, 2017. View at: Publisher Site | Google Scholar
  4. S. Navaratnarajah, C. Han, M. Dianati, and M. A. Imran, “Adaptive stochastic radio access selection scheme for cellular-WLAN heterogeneous communication systems,” IET Communications, vol. 10, no. 15, pp. 1986–1994, 2016. View at: Publisher Site | Google Scholar
  5. Y. Wang, R. Djapic, A. Bergstrom et al., “Performance of WLAN RSS-based SON for LTE/WLAN access network selection,” in Proceedings of the 2014 11th International Symposium on Wireless Communications Systems (ISWCS), pp. 460–464, IEEE, Piscataway, NJ, USA, August 2014. View at: Publisher Site | Google Scholar
  6. A. Habbal, S. I. Goudar, and S. Hassan, “Context-aware radio access technology selection in 5G ultra dense networks,” IEEE Access, vol. 5, pp. 6636–6648, 2017. View at: Google Scholar
  7. M. Lahby, A. Attioui, and A. Sekkaki, “An improved policy for network selection decision based on enhanced-topsis and utility function,” in Proceedings of the 13th IEEE International Wireless Communications and Mobile Computing Conference, IWCMC 2017, pp. 2175–2180, Piscataway, NJ, USA, June 2017. View at: Google Scholar
  8. L.-R. Tang, W.-M. Li, J. Sheng, and B. Fan, “A chaos genetic algorithm based access selection in heterogeneous wireless networks,” Tien Tzu Hsueh Pao/Acta Electronica Sinica, vol. 42, no. 8, pp. 1564–1570, 2014. View at: Google Scholar
  9. N. Blefari-Melazzi, D. Di Sorte, M. Femminella, and G. Reali, “Autonomic control and personalization of a wireless access network,” Computer Networks, vol. 51, no. 10, pp. 2645–2676, 2007. View at: Publisher Site | Google Scholar
  10. F.-C. Yuan, Y.-M. Chen, and T.-Y. Chung, “Intelligent call setup strategy for multimedia communication in heterogeneous wireless network,” Expert Systems with Applications, vol. 39, no. 1, pp. 1298–1305, 2012. View at: Publisher Site | Google Scholar
  11. E. Obayiuwana and O. E. Falowo, “Network selection in heterogeneous wireless networks using multi-criteria decision-making algorithms: a review,” Wireless Networks, vol. 23, no. 8, pp. 2617–2649, 2017. View at: Publisher Site | Google Scholar
  12. M. A. Senouci, S. Hoceini, and A. Mellouk, “Utility function-based TOPSIS for network interface selection in heterogeneous wireless networks,” in Proceedings of the 2016 IEEE International Conference on Communications, ICC 2016, IEEE, Malaysia, May 2016. View at: Google Scholar
  13. S. H. Hashemi, A. Karimi, and M. Tavana, “An integrated green supplier selection approach with analytic network process and improved Grey relational analysis,” International Journal of Production Economics, vol. 159, pp. 178–191, 2015. View at: Publisher Site | Google Scholar
  14. K. Wu, M. Jiang, and H.-Z. Tan, “D2D relay selection based on joint fuzzy and entropy theories with social similarity,” IEEE Transactions on Vehicular Technology, vol. 67, no. 9, pp. 8796–8807, 2018. View at: Publisher Site | Google Scholar
  15. X. Z. Chen, L. H. Zhang, and D. S. Han, “Research on comprehensive performance assessment for wsn based on ANP with equilibrium variable weight and cloud gravity center,” China Communications, vol. 12, pp. 1–9, 2015. View at: Google Scholar
  16. H. Lin, T. H. Pan, H. Wu et al., “Security evaluation of substation communication network based on structure entropy weight and fuzzy comprehensive evaluation method,” Smart Grid, vol. 12, pp. 1174–1179, 2015. View at: Google Scholar
  17. Y. J. Chen, Y. P. Yao, and B. Wang, “Dynamic threat assessment based on structure entropy and IGSO-BP algorithm,” Journal of Systems Engineering and Electronics, vol. 37, pp. 1076–1083, 2015. View at: Google Scholar
  18. S. Zhang, T. Wang, and X. Gu, “Synthetic evaluation of power grid operating states based on intuitionistic fuzzy analytic hierarchy process,” Automation of Electric Power Systems, vol. 40, no. 4, pp. 41–49, 2016. View at: Google Scholar
  19. Q.-T. Nguyen-Vuong, N. Agoulmine, E. H. Cherkaoui et al., “Multicriteria optimization of access selection to improve the quality of experience in heterogeneous wireless access networks,” IEEE Transactions on Vehicular Technology, vol. 62, no. 4, pp. 1785–1800, 2013. View at: Publisher Site | Google Scholar
  20. N. Zarin and A. Agarwal, “A hybrid network selection scheme for heterogeneous wireless access network,” in Proceedings of the 28th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, PIMRC 2017, pp. 1–6, IEEE, Piscataway, NJ, USA, October 2017. View at: Google Scholar
  21. M. M. Hasan, S. Kwon, and J. Na, “Adaptive mobility load balancing algorithm for LTE small-cell networks,” IEEE Transactions on Wireless Communications, vol. 17, no. 4, pp. 2205–2217, 2018. View at: Publisher Site | Google Scholar
  22. F. Khan and M. Portmann, “Backhaul, QoS, and channel-aware load balancing optimization in SDN-based LTE networks,” in Proceedings of the 11th International Conference on Signal Processing and Communication Systems, ICSPCS 2017, pp. 1–10, IEEE, Piscataway, NJ, USA, December 2017. View at: Google Scholar

Copyright © 2019 Chen Zhixiong et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

More related articles

 PDF Download Citation Citation
 Download other formatsMore
 Order printed copiesOrder

Related articles