K-MLP Based Classifier for Discernment of Gratuitous Mails using N-Gram Filtration

Electronic spam is a highly concerning phenomenon over the internet affecting various organisations like Google, Yahoo etc. Email spam causes several serious problems like high utilisation of memory space, financial loss, degradation of computation speed and power, and several threats to authenticated account holders. Email spam allows the spammers to deceit as a legitimate account holder of the organisations to fraud money and other useful information from the victims. It is necessary to control the spreading of spam and to develop an effective and efficient mechanism for defence. In this research, we proposed an efficient method for characterising spam emails using both supervised and unsupervised approaches by boosting the algorithm's performance. This study refined a supervised approach, MLP using a fast and efficient unsupervised approach, K-Means for the detection of spam emails by selecting best features using N-Gram technique. The proposed system shows high accuracy with a low error rate in contrast to the existing technique. The system also shows a reduction in vague information when MLP was combined with K-Means algorithm for selecting initial clusters. N-Gram produces 100 best features from the group of data. Finally, the results are demonstrated and the output of the proposed technique is examined in contrast to the existing technique.