A New Method to Analyze Broadband Internet User Time Preference
Bin Yang, Yang Ge, Yinan Dou, Zhenming Lei · 2012
With the remarkable innovations on network infrastructures, new features of the internet users' behaviors are emerging. Researchers used to analyze the users' log on-line data with statistics and clustering to find out valuable time preference patterns. However, the traditional method to define distance between users would either neglect relevance of time series or lead to high-dimensional crisis which brings difficulties in clustering. In this paper we propose a novel method to identify internet users' time preference patterns. The method not only helps estimate the distance between users' time series but also optimizes DBSCAN in finding out meaningful clusters correspondingly. The experiment results present the method is efficient to deal with the Broadband internet users' log file and also has a good stability with different sampling rate.