Evolutionary Modified Detector Generation Model in Negative Selection Algorithm for Email Spam Detection

Abstract

To deal with the growing problem of unsolicited email in the mail box, a modification of machine learning techniques inspired by human immune system called negative selection algorithm (NSA) is proposed; differential evolution (DE) is implemented to improve the random detector generation in negative selection algorithm. The model is called NSA-DE. The evolutionary algorithm generates detectors at the random detector generation phase of negative selection algorithm. NSA-DE uses local differential evolution for detector generation and local outlier factor (LOF) as fitness function. The theoretical analysis and the experimental result show that the proposed NSA-DE model performs better than the standard NSA.