Socio-Cultural Cognitive Mapping to Identify Communities and Latent Networks

Abstract

Deriving networks and communities from individual and group attributes is an important task in understanding social groups and relations. In this work we propose a novel methodology to derive networks and communities from socio-cultural data. Our methodology is based on socio-cultural cognitive mapping (SCM) and k-NN network modularity maximization (SCM + k-NN) that produces both a latent network and community assignments of entities based upon their socio-cultural and behavioral attributes. We apply this methodology to two real-world data sets and compare the community assignments by our methodology to those communities found by k-Means, Gaussian Mixture Models, and Affinity Propagation. We then analyze the latent networks that are created by SCM + k-NN to derive novel insight into the nature of the communities. The community assignments found by SCM + k-NN are comparable to those produced by current unsupervised machine learning techniques. Additionally, in contrast to current unsupervised machine learning techniques SCM + k-NN also produces a latent network that gives additional insight into community relationships.