Data processing system for detecting health risks and causing treatment responsive to the detection

A method for identifying treatment responsive to a health risk determined from feature data provided by one or more networked data sources, the method comprising:

• generating a feature vector based on a natural language processing (NLP) of input data that is unstructured and representing one or more words provided by a user, with the feature vector comprising one or more features representing one or more health risk factors, the generating comprising:
• applying a set of different NLP models to the input data, each of the different NLP models configured to generate a subset of features for the feature vector;
• selecting the one or more features of the feature vector, from the subset of features of each of the different NLP models, based on cross-validation of different subsets of training data when training machine learning logic;
• classifying, using the machine learning logic that is trained on additional unstructured input data, each of the one or more features of the feature vector to generate a classification metric indicating, for each of the one or more features, that the feature is indicative of a health risk or not indicative of a health risk, the machine learning logic being trained by performing operations including:
• accessing training data that associates words or phrases, of the additional unstructured input data, with a set of features including the one or more features of the feature vector, the one or more features each being labeled with a health risk severity level;
• and
• training the machine learning logic based on the cross-validation of the different subsets of the training data;
• assigning, to one or more of the classification metrics, a prediction weight;
• and
• determining the prediction value for each health risk factor based on the assigned prediction weights.

A non-transitory computer readable medium storing instructions that are executable by one or more processors configured to perform operations comprising:

• generating a feature vector based on a natural language processing (NLP) of input data that is unstructured and representing one or more words provided by a user, with the feature vector comprising one or more features representing one or more health risk factors, the generating comprising:
• applying a set of different NLP models to the input data, each of the different NLP models configured to generate a subset of features for the feature vector;
• selecting the one or more features of the feature vector, from the subset of features of each of the different NLP models, based on cross-validation of different subsets of training data when training machine learning logic;
• classifying, using machine learning logic that is trained on additional unstructured input data, each of the one or more features of the feature vector to generate a classification metric indicating, for each of the one or more features, that the feature is indicative of a health risk or not indicative of a health risk, the machine learning logic being trained by performing operations including
• accessing training data that associates words or phrases, of the additional unstructured input data, with a set of features including the one or more features of the feature vector, the one or more features each being labeled with a health risk severity level;
• and
• training the machine learning logic based on the cross-validation of the different subsets of the training data;
• assigning, to one or more of the classification metrics, a prediction weight;
• and
• determining the prediction value for each health risk factor based on the assigned prediction weights.