The Application Of Stream Data Mining Of Classification In Dynamic Ad Recommender Systems

With the advent of the era of Big Data, information overload starts to be a significant obstacle for user experience. People are starting to think about the value contained in the data gradually. Recommender system is a typical kind of application which people using to mining value in a great deal of data. Recommender system make recommendations to users according to their interests by analyzing users’ behavior and modeling users’ interests.This paper summarizes a few kind of methods commonly used in the recommendation system, such as content-based filtering, collaborative filtering and hybrid filtering. Next this paper give an introduction of two critical technologies (stream processing, stream data mining), which are usually used in real-time analysis. The massive online analysis is now becoming a new research hotspot. Then this paper introduce how the data mining algorithm are used in recommender systems. Finally, we take the advertisement recommender system as an example to give a detailed description of the usage.This papers main work are as follows:we proposed a recommendation way with classification of the features of users and ads. Then we apply the stream classification method very fast decision tree in ads recommendation, which makes the system get the ability of dealing with dynamics. The model can update in real time due to the change of the users behavior. At the end, we integrate the stream processing and stream data mining to design and implement an online dynamic recommender system, which can update model in real time.