Incremental Learning for Spam Detection

Abstract

Through the considerable growth of using email in recent years, the large volume of unsolicited emails called spam made the researchers inspired by text classification techniques to implement systems for filtering such junk emails. Therefore, the goal of current research is to survey all kinds of spams and the issues they cause. It tries to present a method for optimizing spam detection and resolving complexities and diagnostic problems, as well as raising the sentiment of speed and accuracy in spam detection using incremental approach based on collective learning. So this method is to be able to automatically identify spams with mentioned approach. This is because in contrast with batch learning, data is defined as a batch or a group of data at any time which rises the accuracy of spam detection in incremental learning approach. So the goal of algorithm in proposed method is to produce an incremental training model similar to the trained model in batch state. Results of assessments indicate that proposed algorithm can show higher efficiency like other investigated incremental algorithms. Also experiments indicate that proposed algorithm is successful in detecting new input samples through introducing new classes.