Revenue Estimation and Quantification in Sponsored Search Auctions: An Inductive Learning Approach

Abstract

Sponsored Search Auctions (SSA) are major contributors to the search engine’s revenue because of their highly targeted customers and all time available on-line arenas. The Involvement of search users induces a fairly complex dynamics in SSA. It encompasses a gamut of multi-disciplinary research problems starting from modeling users’ clicking behavior to mechanism design. In the proposed work we focus on the users’ response towards advertisements based on the time of query, keywords used in query and position of advertisements. This paper is an effort to estimate and quantify search engine’s pay off using inductive learning which in turn implicitly models users’ clicking behavior and as a byproduct it can help search engine to induce optimality in the auction without sacrificing much of the efficiency of the ranking. Experimental results are presented to demonstrate effectiveness of the proposed scheme.