Voting-Ensemble Classification for Email Spam Detection

On the Internet, e-mail spam continues to become an issue. Spam messages usually contain several duplications of the same text, advertising content and other meaningless documents. Different filtering techniques have been used in previous research to detect such emails. In this study, two phases are proposed. First, various effective traditional machine learning classifiers like Naive Bayes (NB), Support Vector machines (SVMs), k-Nearest Neighbor (k-NN), Decision Tress (DT), Random Forests (RF) and AdaBoost are implemented individually for the detection of spam with accuracy and other performance measures on the dataset. Second, an Ensemble Voting classifier is used, which sums the results of each individual classifier used and calculates the outcome class with the majority votes. Depending on the measures like recall, accuracy and F-score, an efficacy of the dataset is assessed. The research demonstrates that using voting-ensemble classification, the overall accuracy on the SMS Spam dataset improves.