Graph Theoretic Representation and Analysis of Kinship Network Based on Naming System of Kins

Abstract

The kinship network, based on the unique naming system of kins, practiced by the Galo and the Nyishi tribes of Arunachal Pradesh, India, is the focus of this thesis, applying graph theory and social network analysis on genealogical networks and intending to investigate the structural characteristics of these networks using various graph-theoretic parameters. newlineIn the first part of the thesis, we represented the genealogy of the Chiram clan (sub-clan of Nyochi group of clans) of the Galo tribe of Arunachal Pradesh, using the Ore graph, P graph, and Bipartite P graph. On comparative analysis of these representations, P graphs which employ the fewest nodes and arcs, are more appropriate and practical for representing and visualizing genealogical data. Additionally, P graph representation allows for the more accurate and convenient measurement of metrics for social networks and the easier detection of marriage cycles (relinking marriages). newlineIn the second part of the thesis, we represented the genealogy of the Galo tribe (Nyochi group of clans and Heche clan) and the Nyishi tribe (Paate clan), by using P graphs and analyzed the kinship networks on SNA and graph-theoretic parameters. On analysis of kinship networks, it is observed that the networks are acyclic trees with low graph density and no clustering or cliques, with average clustering coieffiecient CC(G) = 0. The networks are divided into many subclans as reflected by the network modularity. The networks obey power law and are scale-free. Centrality measures of the networks reflect the power, prominence, and importance of the individuals of the respective tribes, as well as the number of times the last syllable of the father s name is passed down to their respective siblings. newlineFinally, the thesis examines five well-known community detection approaches on the kinship network of the Nyochi group of clans. On analysis of these algorithms using graph-theoretic parameters, it is found that the Louvain algorithm could be the most effective choice for identifying communities.