US Patent 12137123 Rapid predictive analysis of very large data sets using the distributed computational graph

A system for predictive analysis of very large data sets using a distributed computational graph has been developed. Data receipt software receives streaming data from one or more sources. In a batch data pathway, data formalization software formats input data for storage. A batch event analysis server inspects stored data for trends, situations, or knowledge. Aggregated data is passed to message handler software. System sanity software receives status information from message handler and optimizes system performance. In the streaming pathway, transformation pipeline software manipulates the data stream, provides results back to the system, receives directives from the system sanity and retrain software.