Audience and return recommendation using look alike audience characteristics

Abstract

The system and methods for audience and return recommendation for targeted content items using machine learning and similarity models. The method includes receiving an input dataset comprising of an input audience or specified constraints on the targeted content items. A join operation augments the input dataset by adding individual characteristics data. A trained machine model generates an engagement score for each individual and by using a threshold on the engagement score each individual is classified into a tier category. Based on the engagement level a target audience is generated. Using a similarity model, an expanded audience is generated which increases the size of the target audience by adding similar individuals from the reference audiences. A return is computed using the individuals in the expanded audience while meeting the specified constraints. The expanded audience and return for the targeted content items are returned to the client device.