Recommendation and visualization of similar movies using minimum spanning dendrograms

Exploration of graph structures is an important topic in data mining and data visualization. This work presents a novel technique for visualizing neighbourhood and cluster relationships in graphs; we also show how this methodology can be used within the setting of a recommendation system. Our technique works by projecting the original object distances onto two dimensions while carefully retaining the 'backbone' of important distances. Cluster information is also overlayed on the same projected space. A significant advantage of our approach is that it can accommodate both metric and non-metric distance functions. Our methodology is applied to a visual recommender system for movies to allow easy exploration of the actor-movie bipartite graph. The work offers intuitive movie recommendations based on a selected pivot movie and allows the interactive discovery of related movies based on both textual and semantic features. © The Author(s) 2012.