Design And Realization Of A Recommendation System Based On The Analysis Of Maze Users' Activity

Maze is a network file sharing and transferring system based on P2P, it manages users and file resources through centralized architecture. In this dissertation, we learn how to extract information from user actions and file resources to make clear the relativity between file resources.Firstly, we introduce the status quo of Maze file resources and list the defects of existing recommendation method. We investigate how information of the user actions and the file resources be organized in Maze, and bring forward the feasibility of using such information to judge the relativity between files. We analyze in detail the resource classification information contained in user actions, and conclude that it would bring better recommendation results when using it in constructing recommendation system.After that, we introduce the construction of the relative resources recommendation system, and explain at length the idea in design and realization of important parts of the system, and propose the algorithm to calculate the relativity of file resources.At last we demonstrate the evaluation of this recommendation system through user feedbacks.