Authorship attribution tries to identify the original author of an unattributed text or document. Authorship attribution is a challenging task as it becomes quite difficult to identify original author automatically. Stylometry and authorship recognition or attribution is used interchangeably. Normally authorship attribution is done on the basis of lexical, syntactic and semantic features of a document. More recently, the problem of authorship attribution has gained wide variety attentions in the field of forensic analysis, electronic commerce etc. In this paper various feature selection, reduction and classification techniques are compared for attributing authorship of a document on PAN CLEF 2012 data set. LDA performed 12% well over all other classifiers.