A Modified Naïve Bayes Classifier for Detecting Spam E-mails based on Feature Selection
Argha Ghosh, A. Senthilrajan · 2022 6th International Conference on Intelligent Computing and Control Systems (ICICCS) · 2022
E-mail is the fastest and most formal communicating medium worldwide. However, everyday lots of junk e-mails or spam e-mails get piled up in the mailbox. This research aims to classify spam emails using conventional Naïve Bayes classifier and modified Naïve Bayes classifier, and performs the comparative analysis based on evaluating parameters. This research work proposes the framework for detecting spam emails using a modified NB classifier with Feature Selection. In the name of feature selection, wrapper-based feature subset selection is used as an evaluator and Naïve Bayes-based embedded incremental wrapper subset selection is used as a search method. The conventional Naïve Bayes classifier achieves the accuracy of 87.63% and 79.57% for spam corpus and spambase dataset respectively. Whereas the modified Naïve Bayes classifier attains the accuracy of 93.51% and 89.12% for spam corpus and spambase dataset respectively.