Advanced Spam Filtering In Electronic Mail Using Hybrid the Mini Batch K-Means Normalized Mutual Information Feature Elimination with Elephant Herding Optimization Technique

Abstract

The danger posed by spam is quickly becoming more widespread in today's online environment. It results in a loss of money for both the companies and the users. There have been a lot of clever ideas made to filter out spam. In the first section of this book, the topic of spam is discussed, along with its features, several categories of spam, Spam strategies, various forms of spam, the drawbacks of spam, spam filtering technologies, and the effects of spam letters. When a person is enrolled with social networking sites like Twitter, Facebook, and social job networking sites, spam is more likely to have a negative impact on their online experience. There are four different sorts of spam that may be sent. Usenet spam, instant messaging spam, mobile phone spam, and email spam are the four main types. The term "USENET spam" refers to the practice in which spammers distribute advertising over a large number of newsgroups simultaneously. Spammers make use of instant messaging platforms such as AIM, Windows Live Messenger, and MySpace chat rooms to get user information and then send unwanted messages to those users. The practice of sending unsolicited text messages to those who use mobile devices is known as mobile spam. Because of spam's noisy properties and the time constraints placed on its categorization, determining the optimal spam classification algorithm has become a laborious undertaking in and of itself. The selection of features is a very important part of the classification process since using the most accurate features possible produces the highest accuracy. Optimization techniques such as modified GA, improved RBNN, s-cuckoo search, and enhanced harmony search are introduced with linear, polynomial, and quadratic kernels of SVM for spam classification. This is done in order to achieve a high level of accuracy in spam classification. The Mini batch K-Means Normalized Mutual Information Feature Extraction (KNFE) with Elephant Herding Optimization is used in the first step of the process, which is referred to as feature selection, for the purpose of selecting the pertinent features (EHO). Following the selection of features, a Radial Bias Neural Network classifier will sort the emails into those that are spam and those that are valid. When it comes to the categorization of emails, modified optimization-based feature selection produces superior outcomes than the conventional genetic algorithm. The reproduction process comes after the crossing and the mutation, which is the reason why the improvement is possible. Therefore, there is no possibility of the issue of degradation occurring since the best answer developed by the present generation will be better than those developed in the past. However, genetic algorithms employ a random method to pick parameters. Because of this, they do not perform well in situations when the population size is low, the pace of change is fast, and the fitness function must be selected with great care. It is evident from the findings that the suggested approach to spam categorization was successful in achieving better outcomes.