Use this URL to cite or link to this record in EThOS:
Title: Consistent preference similarity network clustering and influence based consensus group decision making
Author: Kamis, Nor H.
ISNI:       0000 0004 7964 0819
Awarding Body: University of Leicester
Current Institution: University of Leicester
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Access from Institution:
In this thesis, we introduce a novel consensus-based group decision making (CGDM) model by integrating the notions of Social Network Analysis (SNA), clustering and Social Influence Network (SIN). Four main contributions are presented in order to handle a number of issues in CGDM. In dealing with the issue of the consistency of preferences, we introduce a consistency operator and construct a consistency control module for the purpose of securing the correctness of expert preferences. The proposed work guarantees a sufficient preference consistency level for each expert. In the case of inconsistent experts, only minimum changes of preferences are required for them to be consistent, depending on their personal level of inconsistency. The second area of interest focuses on consensus modeling. We develop a novel consensus model by firstly defining the preference similarity network based on the structural equivalence concept. Structurally equivalent experts are partitioned into clusters, thus intra-clusters' experts are high in density and inter-clusters' experts are rich in sparsity. A measure of consensus is defined and the consensus degree of a group of experts obtained reflects the overall agreed solution. A feedback mechanism is presented in dealing with insufficient consensus. We introduce the influence-based feedback system by incorporating the influence score measure in nominating a network leader. Our proposed procedure positively influenced the experts with low consensus contribution to change their preferences closer to each other, by following recommendations from a network influencer. This work guarantees a sufficient consensus level with better clustering solution. Lastly, a procedure of aggregating preferences is laid out whereby the influence function is used in defining a new fusion operator, which helps to aggregate all individual expert preferences into a collective one. This is necessary to ensure that all the properties contained in all the individual preferences are summarized and appropriately taken into considerations.
Supervisor: Chiclana, Francisco ; Levesley, Jeremy ; Gorban, Alexander Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available