Information aggregation in a multi-modal entity-feature graph for intervention prediction

Abstract

A method for providing event-specific intervention recommendations includes identifying, based on an event trigger, one or more sensors, determining, by processing data streams provided by the identified sensors, a set of event related entities and constructing, for each entity of the determined set of entities, an associated feature vector of the entity's features, generating, based on the determined entities and their respective feature vector, an entity-feature-graph, computing, by performing graph classification of a predefined set of allowed interventions based on a trained graph classification model, a ranked list of interventions; and selecting, according to predefined rules, one or more interventions from the ranked list of interventions and providing the selected interventions as recommended interventions for execution. The method can be used to support decision making.