Betweenness measure for social network analysis and image retrieval
Abstract
Betweenness centrality is a measure used for social network analysis. This research applies this centrality measure to image retrieval. The purpose of this thesis is to present an algorithm to compute the betweenness centrality for weighted graphs and compare its performance with the betweenness algorithm for dense unweighted graphs. Betweenness centrality can be used to determine tags' relatedness. It can be used for a multi graph, depicting an image network, to determine which nodes fall on the shortest path computations for a set of nodes in the graph. This in turn, reflects the impact a node has on the graph and its importance in retrieving images using the graph. The research computes the centrality using two data sets - MIRFlickr and ImageCLEF. These data sets are benchmarks. The weighted graph created using these data sets uses a 1-3-9 scale. The thesis presents the results of computation for the betweenness algorithm on the weighted and unweighted versions of the graph. The hypothesis testing showed an increase of 11% in the computation time of the weighted version of the betweenness algorithm as compared to the unweighted version of it. This can be attributed to the fact that the betweenness centrality algorithm determines the shortest paths between nodes using an algorithm that tries to reduce the weights of the paths. The intended impact of the betweenness algorithm for weighted graphs, is to be used with the κ-path centrality algorithm for potential path planning applications. The node importance value determined by the κ centrality values together with betweenness can help determine the optimal paths between nodes.
Degree
M.S.
Advisors
Marshall, Purdue University.
Subject Area
Information Technology
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.