Uso de programação genética para propaganda direcionada baseada em conteúdo

• Ano de defesa: 2008
• Autor(a) principal: Anisio Mendes Lacerda
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Recuperação de informação; Programacao genética; Programação genética (Computação); Analise de conteudo (Comunicação) Processamento de dados
• Link de acesso: http://hdl.handle.net/1843/RVMR-7KWNCX

Internet has become one of the most important media for advertising nowadays. It represents the possibility of global exposure to large audiences at very low cost, which attracts great amounts in investments in advertising. In search advertising, an advertiser company is given prominent positioning in ad lists in return for a placement fee. Because of this, such methods are called paid placement strategies. According to eMarketer's predictions the search advertising market will grow from US$ 16.9Bi in 2006 to US$42Bi in 2011. The most popular paid placement strategy is a non-intrusive technique called keyword-targeted advertising. In this technique, keywords extracted from the user's search query are matched against keywords associated with ads provided by advertisers. The success of keyword-targeted advertising hasmotivated information gatekeepers to oer their ad services in dierent contexts. We refer to the problem of matching ads to a web page that is browsed as content-targeted advertising. In this work, we propose and test a new approach to content-targeted advertising based on genetic programming. Previous work in the literature did not answer important questions such as how to combine the available pieces of evidence or how much importance should be given to each evidence. So, we design a ranking strategy for displaying ads according to their relevance by eectively leveraging all the available evidence. To validate our genetic programming method we performed experiments using a real ad collection and web pages extracted from a Brazilian newspaper. The results obtained show that our genetic programming approach provided gains over a state-of-the-art method of approximately 60% in average precision. Further, the genetic programming was able to learn functions that successfully avoid the placement of irrelevant ads by calculating thresholds based on the page where the ads should be placed. This is very important because of the negative impact of irrelevant ads on credibility and brand of publishers and advertisers. Finally, we perform an extensive and comprehensive analysis of genetic programming individuals in order better understand the results. We realize that there is a great variability between the best genetic programming individuals besides the similarity on the performance of the best individuals. Further, our best genetic programming individualsused only part of all available evidences.