Recommendation System Based On Users’ Behavior Analysis

With the popularity of3G and development of4G, China now is in a period of fastdevelopment of mobile internet. The development of technology makes the informationto be a serious oversupply. While the information carrier like smart phone, pad and smartwear equipment) is becoming smaller and smaller. The information that can be presentedto user by the enterprise is becoming less and less at first glance. All these will lead tothat the user should pay more for what they want. So now a lot of enterprises are payingmore importance to the development of personalized recommendation technology so thatthey will know more about the user. Through this the user will get what they want at lesscost, at the same time the enterprise can hold the user and then form corecompetitiveness.This paper describes the concept, process of data mining. Besides it also describestasks and commonly methods of data mining such as logistic regression, neural networks,cluster analysis and correlation analysis. At the same time, this paper describes thecollaborative filtering algorithm which will be used and the collaborative filteringalgorithms are mainly based on users and items including the theoretical basis andimplementation process. In addition the paper evaluates the collaborative filteringalgorithm including the top-N recommendations and the prediction score and describesthe methods for evaluation.Finally, the paper introduces the framework of recommendation system and use anexample to verify the implementation of collaborative filtering algorithm, whichrecommends jokes to users based on the users’ rating to jokes. We can find collaborativefiltering algorithm can efficiently recommend jokes to users through TPR, PFR andprecision and recall.