Development of composite centrality measures for enhancing the distinguishability of nodes in complex networks using multi-criteria decision making method

Determining the centrality of nodes in complex networks has practical benefits in many areas such as identifying influential nodes, viral marketing, and preventing the spread of rumors. However, there is no consensus on the definition of centrality. Therefore, different centrality measures, such as degree, closeness, and betweenness centrality, have been developed to measure the centrality of a node. However, each centrality measure can highlight various characteristics of nodes in the network from its own perspective. As a result, each centrality measure can rank nodes in a different order in complex networks. In recent years, researchers have focused on approaches that combine multiple centrality measures. In this thesis, we propose a fast and effective method that combines multiple centrality measures using the Analytic Hierarchy Process (AHP) and entropy weighting and Technique for Order Preference by Similarity to Ideal Solutions (TOPSIS) weighting methods for criterion weighting. According to the experimental results obtained by comparing our proposed method with synthetic and real data sets, our proposed method provides competitive results compared to similar studies, while being faster in terms of computation. Thus, our proposed method enables the use of large and dynamic complex networks. For this purpose, analyses were performed on eight different data sets. Betweenness Centrality (BC), Closeness Centrality (CC), Degree Centrality (DC), Eigenvector Centrality (EC), Hubs (HS) and Authorities (HA), and PageRank (PR) analyses were performed for each node. In this thesis, we aimed to increase the distinguishability of nodes by using the centrality results obtained together. By applying the Analytic Hierarchy Process (AHP) from Multi-Criteria Decision Making Methods (MCDM), 127 combinations were obtained for each dataset. In the AHP process, node weights were determined using the entropy weighting and TOPSIS weighting methods. The nodes in each combination were grouped according to their values and normalized to the total number of nodes in the network. Thus, distinguishability ratios of nodes in the network were obtained. In this thesis, it has been observed that the distinguishability of nodes in complex networks can be enhanced by combining different centrality measures associated with the nodes using objective methods.