The Design And Implementation Of Online Advertising System For Search Engine

With the rapid development of Internet technology and the growth of online user community,online advertising has become an irreplaceable source of revenue for Internet companies.Online advertsing for search engine,that is search advertising,occupies the largest scale of market share and achieves the rapidest growth in all kinds of advertisings.The main purpose of search advertising is to manage advertisements triggered by search engine effectively.It aims to constructing appropriate mapping between users and advertisements through analysis on users,meanwhile,promising the profit of search engine and advertisers.Search engine can bring a large amount of users and traffic.Thus it is very essential for companies to convert traffic into revenue.From now on,there is not an open-source online advertising system for sponsored search.It is difficult for some enterprises to gain profit from data generated by search engine effectively.Click-through rate(CTR)prediction is of great importance in search advertising.The performance of CTR prediction algorithtm can have great impact on user experience as well as the profit of advertisers and the search engine.To fulfil requirements mentioned before,this thesis designs and implements an Online Advertising System for Search Engine,and conducts research on the algorithm of click-through rate prediction.The main contributions of this thesis are summarized as the following:(1)This thesis analyzes the functional and un-fucntional requiremetns in accordance with characteristics of search advertising.Incorpating with requirements and key technology of search advertising system,this thesis elaborates overall structure of the system as well as design and implementation of key modules in Online Advertising System for Search Engine.At last,both functional and performance test are performed in this thesis to verify the availability of the system.Key modules of system includes Web Service Module,Advertisement Management Module,Query Processing Module,Advertisement Retrieval Module,Advertisement Ranking Module,Log Collecting and Processing Module.(2)This thesis analyzes drawbacks of a CTR predication algorithm:A Factorization-Machine based Neural Network for CTR Prediction,and proposes a prioritization scheme model for click-through rate prediction of sponsored search,which can exploit the information of continuous and text features better.Experiments show that proposed prioritization scheme can outperform baseline algorithms for CTR predication of sponsored search.