Sponsored ad-based similarity: an approach to mining collective advertiser intelligence

We present a method for mining the intelligence of advertisers to detect product similarities and generate accurate recommendations. In contrast to conventional recommendation algorithms, ours is completely automated and relies solely on publicly available data, namely, the linking of advertisements with Web content. We present a general framework for leveraging linked advertisements to understand object similarity, and provide experimental evidence that the approach yields useful product recommendations.

Citation

Staddon, J. Sponsored ad-based similarity: an approach to mining collective advertiser intelligence. Second International Workshop on Data Mining and Audience Intelligence for Advertising (ADKDD 2008 at ACM SIGKDD '08); 2008 August 24; Las Vegas, NV.