Schema correspondence rule generation using machine learning

Abstract

A schema matching system processes training event data received from multiple sources to determine correspondence rules associating fields in the schemas of each source. To generate the correspondence rules, the schema matching system can use training event data from multiple sources comprising events associated with multiple schemas. Then, based on one or more similarity metrics between data entries of the training event data, the system matches individual events using a machine learning algorithm and, based on the pairs of matching events, corresponding schema fields across the multiple schemas. Based on the matching events and/or user feedback, the schema matching system can generate one or more correspondence rules based on the normalization rules and the corresponding fields of the schemas for later use by one or more transaction monitoring systems on incoming event data.