Weblog Data Structuration
Amine Ganibardi, Chérif Arab Ali · 2018
This paper addresses the issue of Weblog Data structuring within the scope of Web Usage Mining. Web Usage Mining is interested in end-user behavior. Thus, Weblog data need to be cleaned of agent requests and structured per single users and sessions for transaction mining. Since that Webservers record requests interlaced in a sequential order regardless of their source or type, Weblog Data structuring is not trivial when end-users are reticent towards proactive structuration, i.e., authentication or cookies. However, reactive structuration alternatives suffer from quality concerns in terms of session reconstruction. They are agent-centric heuristics that group requests per single pairs of user-agent and IP addresses. This contribution introduces a stream-centric heuristic structuration that drives the session construction to provide relevant single streams before mapping them to single user-agents and IP addresses pairs. The experimental results of the proposed method demonstrate its efficiency in terms of the session construction quality.