The problem of author identification is the following: for a given set of documents having the same author and a new document, we have to decide if this one was written or not by the author of the whole set. For solving this problem we have suggested and implemented various approches : similarity counting, vote technique and supervised learning which explore various model of document representations. The experiences are done using the collections of PANCLEF 2014 challenge and confirmed the interest of our approches and their performance in terms of time execution.