Design And Implementation Of Real-time Personalized Recommendation In Vertical Search

The amount of information is increasing with the rapid popularization and development of Internet.On the one hand,information overload has aggravated the difficulty of information choice,on the other hand,the information provider can not provide high-quality information to users who need.The situation led to the emergence of the search engines,but when the user is unable to accurately describe the keywords that they want to query,search engines also incapable of action.To need this demand,recommendation system generated.As a new information filtering mechanism,recommendation system do not need users to be active,but system predict the interest of users through collecting user behavior.The essence of recommender systems is the connection of users and items through a certain way,and different recommendation system has different contact,which can be through user actions,also through the content feature of the items that can need user interest.Evaluation indexes of the quality of these contacts include the accuracy of recommendation results and real-time of recommend system etc.In this paper,personalized recommendation technology in the vertical search is studied and analyzed,and real-time personalized recommendation system based on the characteristics of vertical search is designed.This paper mainly includes the following content.Firstly,real-time personalized recommendation system architecture is designed,which can feedback in response to users through division of labor and cooperation of offline part and the online part.Secondly,taking advantage of the characteristics of vertical search system,the tags data representing item content features are used for recommendation system,to solve the user data sparse problem.Thirdly,collaborative filtering recommendation algorithm based on graph-model is improved through modifying basing user-item two-nodes graph to basing user-item-label three-nodes graph.Finally this paper realizes and evaluates the design of the real-time personalized recommendation system in vertical search system of video,which is verified to solve the sparse data problem,improve the recommendation accuracy,at the same time ensure real-time of recommendation system.