A comparative analysis of machine learning techniques for spam detection

Abstract

Data Science is an emerging multidisciplinary field whichemploys algorithms, processes, scientific methods to extractinformation and insights in various forms which is bothstructures and unstructured much similar to data mining andprediction analysis. Advertisement and bulk emails, alsocalled as spam, makes an estimate of 62% of the Worldwideinternet traffic. Since 1978, when first unwanted mail wassent, technology have advanced but still the detection ofspams remains a chronophagous and big budget problem inthe field of mathematical sciences. The current study evaluatesthe effectiveness and efficiency of various machine learningtechniques which include K-NN, Decision tree, random forest,Naive Bayes and SVM for spam detection. A data setcomprising of 962 emails containing both genuine emails andspams has been used in this study. Some deep learningtechniques for classification of spams is also suggested forbetter performance.