Dynamic Personalized Ads Recommendation System using Contextual Bandits

Personalized recommendation attracted increasing attention in recent times and had emerged as a viable marketing strategy. The work had used a Click-Through- Rate (CTR) to determine which advertisement to maximise the viewing based on the reward generated. Previous research approaches, which are mostly focused on supervised learning using past data, appear to be yielding decreasing returns. The goal of this research was to deliver individualized dynamic and ongoing recommendations for advertisements. The technique offered in the paper was based on reinforcement learning techniques such as contextual bandits.