The Research Of Precision Advertising System Based On Incremental Learning

With the rapid development of network technology, advertising has become a major means of the network profitable. The growing number of enterprises and organizations come to understand online advertising, and most of them have carried out the online advertising. However, the online advertising takes many forms and possesses dynamic nature. Delivering the online advertising at random makes the users overwhelmed and in a bored mood, so that online advertising runs beyond people's desired. In response to this situation, the concept of precision advertising has been proposed. The Precision advertising is that provide the advertising information which the users may be most interested, and may be looking for at the present time. While through the user's behavior analysis, classify the users intelligently is the effective method to achieve delivering precision advertising. Thus classifying the large-scale users on internet intelligently becomes an important topic for carrying out the precision advertising.As the massive information of network user, utilizing the method data mining can effectively classify the user intelligently. But in view of the information nature real-time and online, which means that users'behavior information is constantly changing under the internet, so that massive information can not be accessed fully for one time. For the data access in batches, the general classification algorithms need update the classification constantly, thus consume a lot of time. Incremental learning process conducts step by step for training dataset, the results of the follow-up learning is based on the previous study. Meanwhile, given the Bayesian classification method can take full advantage of the characteristic prior knowledge to learn, to a certain extent, it can solve the issue of the prior knowledge transmission.Therefore, the article presents an incremental learning algorithm based on Bayesian which is used to classify the users in precision advertising system. It is a continuous and dynamic process that revises the current knowledge utilizing the sample knowledge. By analyzing the behavioral for online users continuously, using the Bayesian classification of incremental to classify the users, and updating the classification again and again, so it can achieve more effectively effect, furthermore, and accomplish precision advertising delivering more effectively. This paper chose the Book-Crossing (BX) dataset as the experimental object of study, which includes 278,858 users (anonymous, but there are demographic information), providing 1149780 million for ratings information in 271,379 books. Through the use of SQL Server 2005 handle the data format into the format we require. Through studying, it indicate that incremental learning can solve the other classifiers'problems with time and energy consuming for real-time online learning, and can obtain better classification results, accordingly, classify the network users accurately and timely, finally reach the purpose of precision advertising. This article also design a precision advertising system, it discusses from system analysis to the implementation of system functions. It mainly present that the user recommended model will be as a common interface. Not only in this system can be applied, other site, as long as the code is embedded in and do the appropriate adjustments, can also be achieved users classified on the current site. The conclusion sums up the entire article and proposes next steps.