Data processing apparatus for accessing shared memory in processing structured data for modifying a parameter vector data structure

A system comprising one or more computers and one or more storage devices storing instructions that are operable, when executed by one or more computers, to cause the one or more computers to:

• access a plurality of risk scoring components associated with an entity;
• map a first set of risk scoring components, from among the plurality of risk scoring components, to one or more physical health categories based on a first model that recognizes risk scoring components as belonging to the one or more physical health categories;
• map a second set of risk scoring components, from among the plurality of risk scoring components, to one or more behavioral health categories based on a second model that recognizes the second set of risk scoring components as belonging to the one or more behavioral health categories;
• generate a plurality of training data structures, wherein each training data structure includes (i) a plurality of first features derived from the first set of risk scoring components mapped to the one or more physical health categories by the first model and (ii) a plurality of second features derived from the second set of risk scoring components mapped to the one or more behavioral health categories by the second model, wherein to generate the plurality of training data structures, the one or more computers are caused to:
• generate a feature vector based on the plurality of first features and the plurality of second features, wherein the feature vector comprises a data field to indicate whether the feature is based on a risk scoring component that is mapped to the one or more physical health categories or is based on a risk scoring component that is mapped to the one or more behavioral health categories;
• assign, for each first feature from among the plurality of first features, a first value of the data field in the feature vector to indicate that a first corresponding risk scoring component is mapped to the one or more physical health categories;
• assign, for each second feature from among the plurality of second features, a second value of the data field in the feature vector to indicate that a second corresponding risk scoring component is mapped to the one or more behavioral health categories if the second corresponding risk scoring component cannot be mapped to the one or more physical health categories;
• obtain, for each training data structure, a target output;
• provide the plurality of training data structures as an input to a machine learning model configured to predict, based on the physical health categories and the behavioral health categories, a risk-level associated with an entity;
• determine an amount of error between the predicted risk-level associated with the entity outputted by the machine learning model and the target output of the plurality of training data structures;
• and
• retrain, based on the determined amount of error, the machine learning model after adjusting one or more parameters of the machine learning model.

A non-transitory computer storage medium encoded with instructions that, when executed by one or more computers, cause the one or more computers to:

• access a plurality of risk scoring components associated with an entity;
• map a first set of risk scoring components, from among the plurality of risk scoring components, to one or more physical health categories based on a first model that recognizes risk scoring components as belonging to the one or more physical health categories;
• map a second set of risk scoring components, from among the plurality of risk scoring components, to one or more behavioral health categories based on a second model that recognizes the second set of risk scoring components as belonging to the one or more behavioral health categories;
• generate a plurality of training data structures, wherein each training data structure includes (i) a plurality of first features derived from the first set of risk scoring components mapped to the one or more physical health categories by the first model and (ii) a plurality of second features derived from the second set of risk scoring components mapped to the one or more behavioral health categories by the second model, wherein to generate the plurality of training data structures, the one or more computers are caused to:
• generate a feature vector based on the plurality of first features and the plurality of second features, wherein the feature vector comprises a data field to indicate whether the feature is based on a risk scoring component that is mapped to the one or more physical health categories or is based on a risk scoring component that is mapped to the one or more behavioral health categories;
• assign, for each first feature from among the plurality of first features, a first value of the data field in the feature vector to indicate that a first corresponding risk scoring component is mapped to the one or more physical health categories;
• assign, for each second feature from among the plurality of second features, a second value of the data field in the feature vector to indicate that a second corresponding risk scoring component is mapped to the one or more behavioral health categories if the second corresponding risk scoring component cannot be mapped to the one or more physical health categories;
• obtain, for each training data structure, a target output;
• provide, the plurality of training data structures as an input to a machine learning model configured to predict, based on the physical health categories and the behavioral health categories, a risk-level associated with an entity;
• determine an amount of error between the predicted risk-level associated with the entity outputted by the machine learning model and the target output of the plurality of training data structures;
• and
• retrain, based on the determined amount of error, the machine learning model after adjusting one or more parameters of the machine learning model.