Comprehensive Data Tree by Actor Messaging for Incremental Hierarchical Clustering

We propose a parallelized and distributed hierarchical clustering system based on Actor Model. In our system, an actor describes a node in the clustering tree, and behavior of the actor receives and processes each incremental inputs containing a data point, based on an algorithm known as BIRCH. Our system adopts a parallel version of CF-Tree and we address inconsistency of asynchronous updating between nodes. As a result, our system can construct and maintain a single large clustering tree on memory across distributed computers with parallel processing of data thanks to actors.