ACTIVITY FORECAST OF THE INTERNET USERS BASED ON THE COLLECTIVE INTELLIGENCE

摘要

The aim of this paper is to compare two modes of activity forecast of the Internet users for the consultation of Web sites. The two modes exploit the phenomena of collective intelligence by using the combined last activity of a group of users as an indicator of the future activity. The first mode, called passive mode, consists in preserving the most popular objects consulted in the past by certain members of the group by hoping that they will be consulted in the future. The second, called active mode, aims at anticipating the download of the objects which correspond to the foreseeable evolution of the group's centres of interest. We present several metric allowing evaluating the effectiveness of the forecast as well as the results of a comparative study relating to 8 days of activity of 2000 users.