Abstract

A method, system, and computer program product for minimizing surprisal context data. The method includes the steps of: identifying characteristics of a data event; receiving an input of rank of at least two identified characteristics of the data event; generating a hierarchy of ranked, identified characteristics based on the rank of the identified characteristics of the data event; and comparing the hierarchy of ranked, identified characteristics to a repository of characteristic context patterns. If at least one reference artifact of the characteristic context pattern matches the hierarchy of ranked, identified characteristics, the characteristic context pattern is broken into pieces, storing the pieces that matched the hierarchy. The stored pieces are then combined into a surprisal context filter pattern with context determined to be probabilistically present within a specified degree of certainty in a data input and compared to a data input of data events to detect anomalous events.