Conceptual clustering on partitioned data: Tree-Weaver

Most knowledge discovery in databases (KDD) research is concentrated on supervised inductive learning. Conceptual clustering is an unsupervised inductive learning technique that organizes observations into an abstraction hierarchy without using predefined class values. However, a typical conceptual clustering algorithm is not suitable for a KDD task because of space and time constraints. Furthermore, typical incremental and non-incremental clustering algorithms are not designed for a partitioned data set. In this paper, we present a conceptual clustering algorithm that works on partitioned data. The proposed algorithm improves the clustering process by using less computation time and less space while maintaining the clustering quality.