Metode Smoothing dalam Naïve Bayes untuk Klasifikasi Email Spam

Abstract

The presence of spam in email lead research on the development of software to classify email spam filter increases. Naïve Bayes is widely used as classification function by spam filter developer. Smoothing method on Naïve Bayes classification function that has been commonly used, namely Add-One smoothing or Laplace smoothing. There are another methods such as Jelinek-Mercer smoothing, Dirichlet smoothing, Absolute Discounting smoothing, and Two –Stage which allegedly able to improve classification accuracy exceeds Laplace smoothing. The experimental results shown accuracy for Naïve Bayes classification function using Laplace smoothing method is 93.72% lower than other smoothing methods which accuration results more than 94%. Naïve Bayes classification function which using Dirichlet smoothing method that gives the best results with accuracy 94.82%.