Hybrid Classifier for Enhancing Accuracy and Performance of Spam and Ham Email Detection

In the contemporary landscape, digital communication reigns supreme, with email being a prominent channel for rapid and widespread information dissemination, also serving as evidence and a promotional tool. Yet, the issue of spam emails imperils data security. Hence, machine learning driven spam classifiers are vital for data preservation. This dissertation rigorously assesses the efficacy of diverse machine learning techniques for spam detection. In response to the exponential surge in spam, devising effective means of identification and filtration is imperative. Leveraging an email dataset, the study compares the performance of Naive Bayes, Support Vector Machines (SVM), Logistic Regression, Random Forest, and K-Nearest Neighbor (KNN) algorithms. This research examines the strengths and limitations of these methods in spam categorization. Evaluation metrics encompassing accuracy, recall, precision, F1 score, and support are considered. Naive Bayes (85%), SVM (85%), Logistic Regression Classifier (84%), Random Forest (84%), and KNN (77% at P=1 & 82% at P=2) are scrutinized, with Naive Bayes and SVM exhibiting notable accuracy. The article contrasts and scrutinizes these five techniques, seeking the most effective spam categorization approach. The findings led to the development of advanced hybrid spam detection systems, amalgamating Naive Bayes and SVM through ensemble technology, promising enhanced protection with the highest accuracy of 87%