Clustering Visual Analysis Of Interpersonal Network Behavior Data In Divide And Recombine Approach

In the era of big data, processing and understanding the large-scale and complicated datasets attracts much attention from both the industrial and academic communities.This paper proposes a novel visual analysis scheme for large-scaled data based on the divide-and-recombine strategy. To classify the users from a mobile user behavior dataset, we first divide the input dataset into a number of subsets. Clustering and other analysis to each subset is then performed individually. Thereafter, the clustering results of all subsets are recombined with a hierarchical clustering algorithm and Ward measure. With a well-designed visual interface, the outliers are interactively identified by the user. A set of visual encoding and visualization techniques are incorporated to visually guide the user to perform the analysis tasks.Our experimental results demonstrate that the divide-and-recombine strategy is a feasible way to deal with big data, which can greatly save resource and time.