Dynamically updated ensemble-based machine learning for streaming data

A system for dynamically updating an ensemble of machine learning classifiers, the system comprising:

• a memory;
• and
• one or more processors communicatively coupled to the memory, the one or more processors configured to:
• receive a first unlabeled data stream;
• provide the first unlabeled data stream as input data to the ensemble of machine learning (ML) classifiers to generate a first prediction,
• wherein the ensemble of ML classifiers comprises a plurality of ML classifiers configured to generate predictions based on input data streams, and
• wherein each set of ML classifiers of the plurality of ML classifiers of the ensemble of ML classifiers includes multiple different types of ML classifiers;
• receive labels for the first unlabeled data stream;
• train an additional set of one or more ML classifiers based on the labels for the first unlabeled data stream and the first unlabeled data stream;
• update the ensemble of ML classifiers by replacing a first set of one or more ML classifiers of the plurality of ML classifiers of the ensemble of ML classifiers with the additional set of ML classifiers, wherein the additional set of ML classifiers replaces an oldest set of ML classifiers of the plurality of ML classifiers of the ensemble of ML classifiers;
• receive an additional unlabeled data stream;
• and
• provide the additional unlabeled data stream as input data to the ensemble of ML classifiers to generate an additional prediction.

A non-transitory computer-readable storage medium storing instructions that, when executed by one or more processors, cause the one or more processors to perform operations for dynamically updating an ensemble of machine learning classifiers, the operations comprising:

• receiving a first unlabeled data stream;
• providing the first unlabeled data stream as input data to the ensemble of machine learning (ML) classifiers to generate a first prediction,
• wherein the ensemble of ML classifiers comprises a plurality of ML classifiers configured to generate predictions based on input data streams, and
• wherein each set of ML classifiers of the plurality of ML classifiers of the ensemble of ML classifiers includes multiple different types of ML classifiers;
• receiving labels for the first unlabeled data stream;
• training an additional set of one or more ML classifiers based on the labels for the first unlabeled data stream and the first unlabeled data stream;
• updating the ensemble of ML classifiers by replacing a first set of one or more ML classifiers of the plurality of ML classifiers of the ensemble of ML classifiers with the additional set of ML classifiers, wherein the additional set of ML classifiers replaces an oldest set of ML classifiers of the plurality of ML classifiers of the ensemble of ML classifiers;
• receiving an additional unlabeled data stream;
• and
• providing the additional unlabeled data stream as input data to the ensemble of ML classifiers to generate an additional prediction.