Research And Application Of User Interest Model In Network Targeting Advertising

With the rapid development of the Internet advertising industry, targetingadvertising, as an emerging online advertising model and technology, is growingrapidly. This mode is subject to much attention of academia and industry becauseit’s accurate, timely, and efficient. The essence of targeting advertising is thepersonalized services technology for advertising resources, delivering advertisingswhich meet user’s interest by analyzing users’ browsing content and behavior, souser interest model is the key to solve the problem of targeting advertising.Currently, the main problems of the research on user interest model contain:(1)Quality problems exist in users’ data collection;(2) Semantic ambiguity problemexist in user interest based on keywords;(3) There are still shortcomings inupdating mechanism of user interest model;(4) The industry is generally lacking oftargeting advertising systemic products or open source projects. To solve theseproblems this thesis is mainly carried out the following work:(1) After analyzing existing data collection methods, a method that collectingeffective browsing content based on browsing behavior is proposed. Research findsthat different browsing behavior has different objective attributes called browsingbehavior feature. VSM is improved and applied to the representation of the userinterest model so as to distinguish the importance between different browsingbehavior features.(2) After analyzing interest model based on keywords, a method thatsemantically extends user interest features based on Wikipedia is proposed. Thismethod calculates the relevance of extended user interest features with thecombination of Wikipedia’s classification system and document link structure andcreates approximate Bayesian Weight Computing Model, improves the accuracy ofuser interest expression.(3) After analyzing distribution pattern of user interest themes, a measure todistinguish between long-term and short-term interest is proposed, and twodifferent update mechanisms are adopted in order to guarantee the quality oflong-term interest while meeting the needs of real-time updating of user short-terminterest.(4) After analyzing the architecture of network targeting advertising, application of user interest model is proposed and parts of key technologies andissues are implemented.Experiment results show that the work done in this thesis effectively alleviatesthe mismatch problem of user and advertisement, improves the effect of targetingadvertising and has a certain application value.