Generative adversarial network recommendation engine

At least one non-transitory machine-readable medium including instructions for a generative adversarial network recommendation engine that, when executed by at least one processor, cause the at least one processor to perform operations to:

• obtain profile data for a user and an organization;
• preprocess the profile data to generate a normalized data set by applying anonymization to sensitive data elements in the profile data, and normalizing values across the profile data;
• train a generative adversarial network using features extracted from the normalized data set including operations to:
• train a generator network to generate synthetic profiles;
• and
• train a discriminator network to distinguish between the synthetic profiles and real profiles including training the discriminator network on real-world data including actual healthcare plan selections, outcomes, and feedback from organizations and employees to distinguish between the synthetic profiles and real profiles, wherein the trained discriminator network evaluates whether suggested healthcare plans are realistic and beneficial for the synthetic profiles;
• generate a set of synthetic profiles using the generative adversarial network;
• calculate risk scores for the set of synthetic profiles based on health and demographic characteristics represented in the synthetic profiles;
• derive a first set of healthcare plan recommendations using the set of synthetic profiles for the organization using aggregate risk scores and aggregate characteristics from the set of synthetic profiles;
• derive a second set of healthcare plan recommendations for the user using individual risk scores and individual characteristics from the set of synthetic profiles;
• determine justification context for each healthcare plan recommendation;
• generate an interactive healthcare plan recommendation user interface, for output on a display device, comprising the first set of healthcare plan recommendations for the organization with corresponding aggregate risk assessments, the second set of healthcare plan recommendations for the user with corresponding individual risk assessments, and the justification context for each recommendation;
• generate a context map for a healthcare plan recommendation of the set of healthcare recommendations;
• generate the justification context for the healthcare plan recommendation using the context map, the justification context including at least one of considerations of the recommendations, pros and cons, or explanations of user or organization attributes used to make a prediction;
• obtain feedback regarding the set of healthcare plan recommendations as an updated feedback dataset, wherein the feedback comprises at least one of user ratings, review information, or subscriber interaction data;
• and
• retrain the discriminator network using the updated feedback dataset by:
• providing the updated feedback dataset as training data to the discriminator network;
• adjusting network parameters of the discriminator network through an adversarial training process where the discriminator network learns to better distinguish between synthetic profiles and real profiles based on the updated feedback dataset;
• and
• iteratively updating weights of the discriminator network until convergence criteria are met to improve accuracy of the discriminator network in evaluating healthcare plan recommendations.

A method for a generative adversarial network recommendation engine, comprising:

• obtaining profile data for a user and an organization;
• preprocessing the profile data to generate a normalized data set by applying anonymization to sensitive data elements in the profile data, and normalizing values across the profile data;
• training a generative adversarial network using features extracted from the normalized data set including operations to:
• train a generator network to generate synthetic profiles;
• and
• train a discriminator network to distinguish between the synthetic profiles and real profiles including training the discriminator network on real-world data including actual healthcare plan selections, outcomes, and feedback from organizations and employees to distinguish between the synthetic profiles and real profiles, wherein the trained discriminator network evaluates whether suggested healthcare plans are realistic and beneficial for the synthetic profiles;
• generating a set of synthetic profiles using the generative adversarial network;
• calculating risk scores for the set of synthetic profiles based on health and demographic characteristics represented in the synthetic profiles;
• deriving a first set of healthcare plan recommendations using the set of synthetic profiles for the organization using aggregate risk scores and aggregate characteristics from the set of synthetic profiles;
• deriving a second set of healthcare plan recommendations for the user using individual risk scores and individual characteristics from the set of synthetic profiles;
• determining justification context for each healthcare plan recommendation;
• generating an interactive healthcare plan recommendation user interface, for output on a display device, comprising the first set of healthcare plan recommendations for the organization with corresponding aggregate risk assessments, the second set of healthcare plan recommendations for the user with corresponding individual risk assessments, and the justification context for each recommendation;
• generate a context map for a healthcare plan recommendation of the set of healthcare recommendations;
• generate the justification context for the healthcare plan recommendation using the context map, the justification context including at least one of considerations of the recommendations, pros and cons, or explanations of user or organization attributes used to make a prediction;
• obtain feedback regarding the set of healthcare plan recommendations as an updated feedback dataset, wherein the feedback comprises at least one of user ratings, review information, or subscriber interaction data;
• and
• retrain the discriminator network using the updated feedback dataset by:
• providing the updated feedback dataset as training data to the discriminator network;
• adjusting network parameters of the discriminator network through an adversarial training process where the discriminator network learns to better distinguish between synthetic profiles and real profiles based on the updated feedback dataset;
• and
• iteratively updating weights of the discriminator network until convergence criteria are met to improve accuracy of the discriminator network in evaluating healthcare plan recommendations.