Long Short-Term Memory Networks for Email Spam Classification

In this internet era, email becomes very popular because of its fastest and easiest way of sharing information with others. Spam is the major problem, which sends unwanted messages to huge recipients to steal users' confidential information, spread malware, and advertise any products. Even though most Machine Learning algorithms exist to detect spam emails, none of them predicts spam emails accurately. The input layer, hidden layer, and output layer of neural networks are used to categorize spam emails. However, there is no memory for it to keep past data. RNNs are one variety of neural networks which remembers past data but suffers from the problem of vanishing gradient and exploding gradient. To overcome this disadvantage, the proposed system uses Long Short-Term Memory to classify the spam and ham emails using Spam base and Ling Spam dataset. It keeps track of the email's previous information and decides which information is relevant for classifying spam emails and ignores irrelevant information. LSTM outperformed well with an accuracy of 97.4% when compared with other Deep Learning techniques.