Content-based Dynamic Email Spam Detecting Using Fuzzy Granular Computing Approach

Spam detection is a significant problem which is considered by many researchers by various developed strategies. The best and main spam detection technique should consider and scan the content of the messages to find spam. This research concerns the development of the certain category of granular computing as a classifier for spam detection. In this research, Fuzzy Granular Computing Classification Algorithm (F_GrCCA) will be used beside the clustring algorithm to provide powerful framework for classification of emails as spam or non-spam. In fact, the core structure of this framework will be constructed by selected clustring algorithm of data mining. Then, in the secondary structure F_GrCCA will be used to reduce the likelihood of the classification errors at the area of high overlap between the classes and cover a substantial number of patterns which do not belong to any cluster. In this study, the performance of proposed technique will be measured by several evaluation criteria as: spam recall (Rs), spam precision (Ps), accuracy which deals with false positive, false negative and FB measure. The effectiveness of fuzzy granular computing in spam detecting will be analaysed and compared with Naive Bayes Classifier according to achieved accuracy. These algorithms were trained and tested on a set of 4601 email messages in which 1813 were spams and 2788 were non-spams. The experiments will be performed based on different training set size and extracted feature size.