Abstract

Abstract: Spam email has become a major concern for customers and Internet providers alike. One of the main obstacles to its removal is that the proposed remedies need to have a very low, practical, false-positive rate. Natural Language Processing is a major subject of research in a range of areas. Text Classification is part of the NLP in which a range of approaches are used to transform text into a machine-readable format. Methods for categorizing texts include tokenization, speech-part tagging, stemming, and chunking. Using Scikit-Learn Classifiers to train the model to recognize spam and ham communications following the implementation of these data processes, offers us a categorized data set in which the model is trained to detect spam and ham messages. By exploring and comparing the relative capabilities of several machine learning approaches, we constructed a model for the problem of spam or ham transmissions. We are looking at the problem of classification in the context of naive Bayes, one of the most utilized machine learning models in the field of spam filtering. The performance metrics of the algorithms we employed in this study are comparable logically by Naive Bayes (NB) and Multinomial Naive Bays (MNB). The approach we offered in our 'Spam Mail Collection' data set showed an average accuracy of 98.49% utilizing the Naive Bayes (NB) model. The suggested system would compare the contents of the message with the contents of the spam keyword database, and the mail will be categorized as spam, if the contents match the contents of the database.Keywords: Naive Bayes (NB); Multinomial Naive Bayes (MNB); part-of-speech (POS);Natural Language Processing(NLP); Confusion Matrix(CM).