\(\alpha \)-Based Similarity Metric in Computational Advertizing: A New Approach to Audience Extension

Over the past few decades, advertising has undergone significant evolution, with online advertising now the most widely used form to reach potential audiences globally. Advertisers face the challenge of targeting the right audience through media channels while working within limited budgets. However, campaigns often attract small audiences, which has led to extensive research into inducing preferable attributes from campaign data to reach a wider range of customers. Audience expansion techniques offer a promising solution to identifying potential audiences with similar characteristics to the seed users, who are likely to achieve the business goal of a targeted campaign. Typically, the ultimate goal is to achieve the maximum impressions possible at a certain cost per thousand impressions. In this paper, we propose a distance-based approach that uses a hyperparameter to compute the weighted average to find the nearest neighbors of a target campaign from the historical dataset (seed audience). This approach will be used to determine the total impressions and cost per thousand impressions. To extend our potential audience, we will use heuristic measures to find the best set of features to render the maximum impressions at a reasonable cost per thousand impressions [1].