Proceedings of the 39th International Academic Conference, Amsterdam




This paper aim is to study the process of real time bidding or real time auctions for online digital advertising. Real time bidding drives the focus of bidding strategy from the user's profile by calculating a bid for each impression in real time. Real Time Bidding uses computers and multiple software’s which implements multiple algorithms to display ads per impression via real time auction. It has been seen that by taking different parameters (e.g. conversion rates for a targeted audience), those account for varied prices at different market segments or pricing schemes. The data mining model implemented is the Statistical Arbitrage Mining (SAM). The campaigns use the CPA (cost per action) method on the meta-bidder to accomplish CPM (cost per mille-impressions) ad inventories paradigm thereby reducing the advertiser’s risk. In SAM, trying to seek the optimal bidding price to maximize the expected arbitrage net profit is the net goal. A modern portfolio base is implemented to manage the risk. The Expectation - Maximization (EM) fashion is used to estimate the profit of each campaign and thereby maximize it. By using this, the meta-bidder successfully catches the statistical arbitrage opportunities in RTB. Also using the concepts of finance, the calculation of risk is done for each campaign.

Keywords: Expectation - Maximization; Bidding; Statistical Arbitrage Mining

DOI: 10.20472/IAC.2018.039.042

PDF: Download

Copyright © 2024 The International Institute of Social and Economic Sciences,