Apparatus and methods for expanding clinical cohorts for improved efficacy of supervised learning

A method for machine-learning using a seed data set and an unstructured data set to train a supervised machine-learning model, wherein the method comprises:

• receiving, by at least a processor from a database, at least a seed set of labeled data, wherein the at least a seed set of labeled data contains information and classification methods, wherein the information is structurally analogous to information required in data to be classified, wherein the at least a seed set of labeled data comprises electronic health record data,
• using a neural network;
• identifying and abstracting, by an auxiliary feature extraction module, first features associated with the at least a seed set of labeled data, wherein the first features are associated with at least one of:
• training and determining embeddings;
• extracting, by a task-specific feature extraction module, in response to the auxiliary feature extraction module, second features from the at least a seed set of labeled data, wherein the second features are associated with a machine-learning task associated with the neural network;
• wherein feature extraction is associated with identifying non-domain specific features and isolating such features for processing;
• initializing the supervised machine learning model using the second features;
• training, by the at least a processor, an embedding model based on the at least a seed set of labeled data and the first features, wherein training the embedding model comprises:
• adjusting connections and weights between nodes in adjacent layers of the embedding model;
• determining, by the at least a processor and using the embedding model, a plurality of first vector representations corresponding to the at least a seed set of labeled data;
• mapping, using the at least a processor, the plurality of first vector representations and the plurality of second vector representations within a shared vector space;
• determining, by the at least a processor and using the embedding model, a plurality of second vector representations corresponding to an unlabeled set of data;
• identifying, by the at least a processor, at least one first data set from the unlabeled set of data as having a predetermined distance to at least one second data set from the at least a seed set of labeled data based on the respective vector representations of the at least one first data set and the at least one second data set within the shared vector space;
• assigning, by the at least a processor, labels to the at least one first data set with same labels associated with the at least one second data set;
• increasing, by the at least a processor, a size of a training data set for the supervised machine-learning model by providing the at least one first data set and the labels as labeled training data to the supervised machine-learning model;
• and
• training, by the at least a processor, the supervised machine-learning model using the labeled training data to increase an accuracy of the supervised machine-learning model.