Introduction
Related work
Article | Uses a network? | LSGDM | Description |
---|---|---|---|
Urena et al. [5] | ✓ | A social network is created based on similarity, confidence, and consistency. The main idea is that some opinions with high similarity, confidence, and consistency have a greater influence on the opinions of others | |
Chao et al. [9] | ✓ | A clustering method is used to detect non-cooperative behavior, and a weighting process is used to manage this behavior. The number of clusters should be determined, which is different for different group-decision scenarios | |
Liao et al. [11] | ✓ | \(K\)-means is used for clustering, which requires the \(k\) parameter initialization | |
Du et al. [12] | ✓ | Opinion punishment and weight punishment are used to manage non-cooperative behaviors. The number of clusters should be determined which is different in various group decision scenarios | |
Wu et al. [15] | ✓ | A trust network was used to estimate the unknown preference values and experts’ weight determination | |
Wu et al. [16] | ✓ | A trust network was used to estimate the unknown preference values and extract the reputation between experts as historic actions | |
Taghavi et al. [17] | ✓ | A feedback-based influence network was proposed, in which the influence between agents was calculated by trust, self-confidence, and similarity | |
Herrera-Viedma et al. [18] | Using the additive-consistency concept, a procedure is provided to estimate the missing information in an expert’s incomplete preference | ||
Cheng et al. [19] | ✓ | A weight allocation method is provided by analyzing the tie strength and topology structure of social networks | |
Xu et al. [20] | Prospect theory was used to solve the group decision-making problem | ||
Zhang et al. [21] | Heterogeneous preference structures were accepted as inputs | ||
Wu et al. [22] | ✓ | \(K\)-means was used for clustering and clusters were allowed to change. To use \(K\)-means, the parameter \(k\) should be determined. Changing the clustering approach has computational overhead in LSGDM | |
Lu et al. [23] | ✓ | ✓ | \(K\)-means is used for clustering. To use \(K\)-means, the parameter \(k\) should be determined |
Wu et al. [24] | ✓ | ✓ | \(K\)-means is used for clustering. To use \(K\)-means, the parameter \(k\) should be determined |
Trillo et al. [25] | ✓ | NLP techniques were used to detect the degree of positivity and aggressiveness of experts. The main idea revolves around sentiment analysis | |
Zhong et al. [26] | ✓ | A combination of similarities in the evaluation information is used for clustering. The K-means algorithm is used for clustering. To use \(K\)-means, the parameter \(k\) should be determined | |
Liu et al. [27] | ✓ | The probabilistic K-means clustering algorithm is used to improve the selection of the initial centroids. However, the \(k\) parameter should be determined | |
Dong et al. [28] | ✓ | Leaders and their followers are detected in the network. Leaders influence the opinion of their followers | |
Zhang et al. [29] | ✓ | A feedback mechanism is provided for each expert by considering the leadership and bounded confidence levels of experts | |
Chu et al. [30] | ✓ | ✓ | A two-stage consensus-reaching method is proposed in which cluster preferences can change |
Ding et al. [31] | ✓ | ✓ | A negative conflict relationship is considered between DMs. Feedback is generated for each DM which may not be suitable in LSGDM |
Wu et al. [33] | ✓ | Trust-based recommendation mechanism is applied | |
Gavrilets et al. [35] | ✓ | The dynamics of consensus building in groups is investigated which is composed of individuals who are heterogeneous in preferences and have different personality traits (agreeability and persuasiveness) and reputation | |
Li et al. [36] | ✓ | Extracts the influence network from expert opinions and social networks | |
Zhang et al. [38] | ✓ | A signed network is used, in which both positive and negative relations can be considered | |
Triantaphyllou et al. [39] | ✓ | A post-consensus analysis was used in the log file to identify any dynamics that may exist in the way experts make ranking decisions | |
He et al. [40] | ✓ | ✓ | The shadowed theory is used for preference presentation and clustering. The network construction complexity is O(\({n}^{2}\)) as it requires pairwise analysis |
Xiao et al. [42] | ✓ | Centrality, consistency, and similarity indices were used for weighting experts. Using similarity makes the complexity O(\({n}^{2}\)) as it requires pairwise analysis | |
Zhang et al. [43] | ✓ | ✓ | Leadership and non-cooperative behaviors were detected in the trust network. A social network is needed for the input |
Li et al. [44] | ✓ | A fast unfolding algorithm was used to reduce the dimension of the large-scale DMs and the experts’ weights were obtained by social network analysis techniques. This approach needs a social network as input | |
Liao et al. [46] | ✓ | ✓ | Two different roles are considered |
Chao et al. [47] | ✓ | A two-layer network was used with an inner layer consisting of participants whose preference similarities and trust relations were known. The outside layer includes participants whose trust relations cannot be determined. The complexity of the proposed method is approximately O(\({n}^{3}\)) | |
Li et al. [48] | ✓ | An interaction network is used to detect and manage manipulative behaviors | |
Xiong et al. [49] | ✓ | ✓ | A clustering method with historical data to support large-scale consensus-reaching process |
Article | Pros and cons | ||
---|---|---|---|
Selection approach | |||
All | [35] | No need for DM identification for suggestions One rule for all DMs Simple | |
Subgroup | Accelerates the consensus process as the preferences change within a group Needs more calculations for subgroup detection and subgroup suggestion | ||
Decision-maker | Slows down the consensus process as the preferences change one by one Used in small groups | ||
Adaptive | Speeds up the consensus process as the preferences change within a group at first Needs more calculations for subgroup or DM selection | ||
Suggestion generation approach | |||
Collective opinion | A proper aggregation function is needed | ||
Opinion dynamics | The network’s existence is mandatory | ||
Network-based feedback | The network is created | A network is created based on the input information | |
There is a predefined network | It is assumed that additional information is available from another network |
Background knowledge
Preference relations
Single-valued neutrosophic preference relations
Consistency of decision-maker’s preference
Consensus level
Graphs and networks
Problem statement
The proposed approach
Preprocessing
Opinion transformation
The complement process
Calculating the decision-maker’s certainty
Determining the preference’s class
Creating the class graph/network
Creating the two-layer network
Consensus process
Feedback
Selection process
Evaluations
Simulation scenarios
Simulating a group decision-making scenario
Number of decision-makers | Number of alternatives | Possibility of a fully consistent preference | Possibility of a certain preference |
---|---|---|---|
1000 | 4 | 1 | 0.1 |
Simulation with different numbers of decision-makers, alternatives, and similarity measures
-
It can be concluded that if the number of decision-makers is larger than \(\left(m\right)!\) the number of decision-makers does not have a significant effect on the number of iterations of the algorithm. In that case, the structural features of the class’s network and the number of preferences belonging to a class determine the number of iterations. The proof of this is shown in Appendix A.
-
In some diagrams (e.g., simulation number 32), a decrease in the proposed similarity occurred in the consensus process. This decrease occurs because the selected nodes for integration at this stage have a lower eigenvector. In other words, the population or certainty of some classes was high enough to overcome the eigenvector criterion, and the preferences shifted to classes with more population or more certainty, rather than moving to more similar classes.
-
In most diagrams, the degree of similarity in the final stage changes sharply. This is because in the early stages, nodes that have a smaller population are integrated and their changes do not cause a significant change in the similarity. On the other hand, with the migration of nodes in each stage, the population of nodes in the final stages increases, and consequently, the number of links that change in further stages increases. Another reason is the dissimilarity in preference data. Data generation is performed in such a way that it is produced from any kind of consistent and certain class, which means that there are nodes in the network that have no similarities, and the preferences may converge to two opposite preferences. Therefore, the latest change is for the migration of the last two opposite classes which sharply increases the similarity.
-
As it turns out, at the end of the process, when the class of preferences is the same, the Euclidean similarity is still low. Therefore, it can be concluded that the use of Euclidean similarity is not a suitable measure for the problems with the goal of ranking the alternatives where there is no need for Euclidean similarities to be high. In other words, the low Euclidean similarity between preferences does not mean that decision-makers have not reached a consensus on the ranking of available alternatives. Although Cosine distance has been used in fewer studies, it provides better results than Euclidean distance.
Comparisons
Preferences data | Similarity changes in each iteration | Parameters | |
---|---|---|---|
[22] | Number of decision-makers | 20 | |
Number of alternatives | 4 | ||
[5] | Number of decision-makers | 25 | |
Number of alternatives | 3 | ||
[48] | Number of decision-makers | 50 | |
Number of alternatives | 4 | ||
[11] | Number of decision-makers | 20 | |
Number of alternatives | 4 | ||
[49] | Number of decision-makers | 20 | |
Number of alternatives | 5 |