The disputed federalist papers
Glenn Fung · 2003
In this paper, we use a method proposed by Bradley and Mangasarian "Feature Selection via Concave Minimization and Support Vector Machines" to solve the well-known disputed Federalist Papers classification problem. We find a separating plane that classifies correctly all the "training set" papers of known authorship, based on the relative frequencies of only three words. Using the obtained separating hyperplane in three dimensions, all of the 12 disputed papers ended up on the Madison side of the separating plane. This result coincides with previous work on this problem using other classification techniques.