Methods, apparatuses and computer program products for generating predicted multi-drug contraindication data objects

A computer-implemented method for generating predicted contraindication data objects comprising:

• generating, by one or more processors, a plurality of reduced drug input vectors based at least in part on a plurality of combined drug input vectors, wherein a combined drug input vector of the plurality of combined drug input vectors corresponds to a first drug and a second drug;
• inputting, by the one or more processors, a plurality of record data objects and the plurality of reduced drug input vectors to an encoding layer of an interaction-attentive machine learning model to receive a plurality of encoded multidimensional tensors;
• inputting, by the one or more processors, the plurality of encoded multidimensional tensors to the interaction-attentive machine learning model to receive an interaction-attentive prediction data object, wherein the interaction-attentive prediction data object describes a predicted interaction between the first drug and the second drug;
• inputting, by the one or more processors, the plurality of record data objects to an interaction-inattentive machine learning model to receive an interaction-inattentive prediction data object, wherein the interaction-inattentive machine learning model comprises a logistic regression model that is trained to determine a predicted outcome based at least in part on the plurality of record data objects;
• and
• training, by the one or more processors, the interaction-attentive machine learning model based at least in part on a predicted significance indicator corresponding to a combination indicator that indicates a likelihood that a defined outcome is associated with a combination of one or more of the plurality of combined drug input vectors, wherein the likelihood is determined based at least in part on a comparison of the interaction-attentive prediction data object and the interaction-inattentive prediction data object.

One or more non-transitory computer-readable storage media including instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:

• generating a plurality of reduced drug input vectors based at least in part on a plurality of combined drug input vectors, wherein a combined drug input vector of the plurality of combined drug input vectors corresponds to a first drug and a second drug;
• inputting a plurality of record data objects and the plurality of reduced drug input vectors to an encoding layer of an interaction-attentive machine learning model to receive a plurality of encoded multidimensional tensors;
• inputting the plurality of encoded multidimensional tensors to a transformer layer of the interaction-attentive machine learning model to receive an interaction-attentive prediction data object, wherein the interaction-attentive prediction data object describes a predicted interaction between the first drug and the second drug;
• inputting the plurality of record data objects to an interaction-inattentive machine learning model to receive an interaction-inattentive prediction data object, wherein the interaction-inattentive machine learning model comprises a logistic regression model that is trained to determine a predicted outcome based at least in part on the plurality of record data objects;
• and
• training the interaction-attentive machine learning model based at least in part on a predicted significance indicator corresponding to a combination indicator that indicates a likelihood that a defined outcome is associated with a combination of one or more of the plurality of combined drug input vectors, wherein the likelihood is determined based at least in part on a comparison of the interaction-attentive prediction data object and the interaction-inattentive prediction data object.