Within the community detection problem it is possible to use either the structural dimension or the composition dimension of the social network; on the first case the communities contain groups of well-connected and dissimilar nodes whereas on the second case, the communities contain groups of similar but loosely connected nodes. Therefore the amount of information extracted is reduced as one of the dimensions is discarded. The objective of this Thesis is to propose a novel approach for detecting communities in which the structural and composition dimensions are integrated in such a way the communities contain groups of well-connected and similar nodes. This approach requires first, a new definition of community that includes both dimensions of the network, then a new community detection model suited for this new definition that allows us to find groups of well-connected and similar nodes. The model starts introducing the notion of point of view that allows the division of the composition dimension for analyzing the network from different perspectives. Then the model influences the community detection process by integrating the composition information into the graph structure. The last step is the social network visualization that places the nodes according to their structural and compositional similarities and that allows us to find important nodes regarding the interaction between communities.
