US Patent 10515095 Detecting clusters and relationships in large data sets

From a directional acyclic graph (DAG) corresponding to a data volume, a matrix of relationships of a first node and a second node is computed. The DAG includes a set of nodes and a set of directional edges, the set of nodes including the first node and the second node. An inverted matrix is computed from the matrix, which retains values in only those column positions where a value in a first row corresponding to the first node is opposite of a value in a second row corresponding to the second node. From the first row, a first cluster of nodes related to the first node is constructed by selecting nodes corresponding to those columns where the first row has a value indicating an affirmative relationship. Information representing the first cluster of nodes a second cluster of nodes are output.