Research On The Optimal Strategies For Sponsored Search-Based Keyword Advertising Services

Search-based keyword advertising, also known for sponsored search, is a kind of online marketing tools using search engine. It is an important online marketing tool not only providing major revenue stream for search engine but also acting as a great sale channel for many businesses. This special advertisement meets both the requirement of the information need of internet users and the advertisement need of businesses especially small businesses. It enables them to promote their products to customer directly. Compared with traditional advertisements, search-based keyword advertising has the following merits: strongly targeted, lower cost, budget controllable, easy operation, distinguishable and measurable advertising effect etc.Research related keyword advertising from the point of view of advertisers is be in leading strings, most of them focus on how to bid for a specify keyword such as to maximize profit of advertisers. The domains about how to make budget allocation over time and how to choose keywords and how to react on the opponent's strategy, though is very important for advertisers, remain untouched. In order to narrow the gap between current research and real world practice, this dissertation presents two research topics about keyword advertising. The first topic takes the angle of the macro-level, which is about how to allocate budget under different ranking and pricing mechanisms adopt by search engines. The second topic takes the angle of the micro-level, which is about how to choose the most profitable keywords and the optimal competitive strategies. The content of this dissertation includes:Firstly, we build a budget allocation optimal control model for single search engine. Based on some reasonable assumptions about keyword advertising we build the dynamic optimal model. There are two aspects of this topic. First, we consider the scenario when advertisers run advertisement on search engine which using the pure generalized second price mechanism and we present optimal budget allocation strategies under the different trend of potential clicks of search engine. Second, we add the quality score impact to the generalized second price model to cope with the situation when advertisers run advertising on quality-based generalized second price mechanism search engine. By making the reasonable assumption that the quality score is the increasing function of click through rate, we present the optimal budget allocation policies with different circumstances.Secondly, we build optimal budget allocation model for multiple search engine. It is about how to allocate resource among different search engines. We consider the following situations: optimal budget allocation policies across generalized second price search engines; optimal budget allocation policies across quality-based generalized second price search engines; optimal budget allocation policies across search engines with generalized second price and quality-based generalized second price simultaneously. We propose different optimal budget allocation strategies for these three situations. In order to evaluate the propositions, we build a simulation tool to evaluate the proposed policies.Thridly, we study the optimal keyword choosing policies at the micro-level. Define the keyword choosing problem, present the optimal policies when only single keyword could be used at each period including the length of keyword list is 2 and larger than 2. We present four new old keyword-choosing policies since Gittins Index cannot be computed for new keywords. We also study the problem when multiple keywords can be used at each period and present the policies, including the following scenario: choosing policies when the returns of keywords are normal distribution; choosing policies for keywords without assumption of return. By simulation approach we evaluate the solutions under different setting and show that none of these solutions are necessary optimal, advertiser needs to choose solution according to the specify environment. We also build up the keyword-choosing model and present solution with perfect informationForthly, take the competition of advertisers'point of view, we research the optimal bidding strategies of advertisers by using game theory. First, we define and abstract the slot competition between advertisers, which is the essential step of building up game model, and then we present optimal strategies of single-period and multi-period respectively.Finally, we analysis and design the Artificial-Computation-Parallel framework-based decision support system for keyword advertising. We build a keyword advertising experiment platform base on the analysis of artificial system, computational experiment and parallel execution of keyword advertising. Furthermore, the technical framework, content framework and the decision support system framework of the experiment platform are carefully designed. In order to realize the platform, we design the scenes and bidding agent of the experiment platform.Results of this dissertation to some extent make up the research gap for the optimal budget allocation and keyword selection and optimal bidding in keyword advertising, and are useful references for business to participate in the keyword advertising.