Efficient Email Classification Algorithm for Better Customer Support

Abstract

Email has become one of the most important daily applications. Due to the continuing rise in email users, the number of unsolicited emails, also known as spam messages, has increased enormously. It is a big challenge to handle and classify this huge number of emails, while in spam message separating, traditional technologies are quite effective. Various steps must be taken to make spam filtering more accurate. We worked on this effort to identify and filter spam messages while delivering them. The high dimensionality of email through syntactic feature selection was addressed in most methods presented for resolving this issue. This article deals with an efficient approach to email filtering based on semantic techniques and similarity measures for dropping the huge number of extracted textual features, and therefore, the space and time complexities are reduced. This study is intended to enhance the manually defined rules. The algorithms currently used for the use of machine learning in a telecommunications company's business. The suggested model with long short-term memory (LSTM) retains data in the memory for a longer duration of time and has a better F1-score and accuracy. It integrates or migrates the administrative and maintenance tasks from a manually designed rule-based model into a machine learning model. It should also increase the flexibility of the model.