An Empirical Study of a Simple Naive Bayes Classifier Based on Ranking Functions
Kinzang Chhogyal, Abhaya Nayak · Lecture notes in computer science · 2016
Ranking functions provide an alternative way of modelling uncertainty. Much of the research in this area focuses on its theoretical and philosophical aspects. Approaches to solving practical problems involving uncertainty have been, by and large, dominated by probabilistic models of uncertainty. In this paper we investigate if ranking functions can be used to solve practical problems in an uncertain domain. In particular, we look at the problem of identifying spam e-mails, one of the earliest success stories of probabilistic machine learning techniques. We show how the probabilistic naive Bayes classifier can easily be translated to one based on ranking functions, and present some experimental results that demonstrate its efficacy in correctly identifying spam e-mails. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.