Research On The Portfolio Optimization Of The Online Keyword Advertising

In recent years, online keyword advertising has become one of the effective promotion methods for many companies and is developing very fast with the development of the Internet. Keyword advertising is becoming the focus globally. But the current research is mainly in making advertising bid strategy on a single platform. Actually more and more companies start to do business through several different platforms. So how to make portfolio decisions among platforms is a theory expansion and also important to those advertisers.Starting with the current situation and characteristics of keyword advertising, this paper introduces three major advertisement providers and analyzes the importance to allocate resources among different platforms. Then based on the Bayesian Network Theory and Markowitz investment portfolio theory, this paper establishes ROI forecasting model for each platform and the portfolio optimization model. What’s more, a comprehensive management system is also researched in this paper. Following are the main contents of this paper.Firstly, this paper analyzes keyword advertising’s history, development, current situation, achieved research result, some key players to provide this kind of service and the difference among them. Meanwhile some basic background theoretic knowledge is also introduced.Secondly, it analyzes how to set up ROI forecasting model through Bayesian Network Programming for different platforms. Then it also double checks the accuracy of the established model.Thirdly, this paper establishes a quantitative model for portfolio optimization based on the ROI forecasting and Markowitz theory. After this, it also validates the model using real business data.Fourthly, it suggests a management system for companies’online keyword advertising in a macro view and analyzes the risk situation in doing this kind of advertising.This paper uses the Bayesian Network Package written by Kevin Murphy-BayesNet Toolbox together with Matlab programming to set up the Bayesian Network forecasting model.This paper establishes a new framework to conduct keyword advertising portfolio decision among different kind of platforms based on theoretic knowledge and programming tools. Meanwhile, the validity and accuracy has been checked by real business data. It not only expands the research on keyword advertising, but also provides theoretic reference for the advertisers to make portfolio decision.