BibTeX:

@article{IJIRSTV2I1041,
     title={Optimized Incremental SVM based Classifier for Spam Filtering using Internet Acronyms},
     author={Indrajeet Singh Jhala and Dr. Pankaj Dalal},
     journal={International Journal for Innovative Research in Science & Technology},
     volume={2},
     number={1},
     pages={163--176},
     year={},
     url={http://www.ijirst.org/articles/IJIRSTV2I1041.pdf},
     publisher={IJIRST (International Journal for Innovative Research in Science & Technology)},
}

The word Spam as applied to email means Unsolicited Bulk Email. Unsolicited means that the Recipient has not granted verifiable permission for the message to be sent. Bulk means that the message is sent as part of a larger collection of messages, all having substantively identical content. A message is Spam only if it is both Unsolicited and Bulk. Spam is a key problem in electronic communication, including large-scale email systems and the growing number of blogs. The anti-spam community has been divided on the choice of the best machine learning method for content-based spam detection. In many traditional machine learning applications, SVMs are applied in batch mode. That is, an SVM is trained on an entire set of training data, and is then tested on a separate set of testing data. Spam filtering is typically tested and deployed in an online setting, which proceeds incrementally. In this paper, an online incremental SVM is proposed which classifies emails as spam or ham (normal mail). The traditional textual features are augmented with character level Features, thus, extending the feature set used to train the SVM. This approach exploits the fact that most of the user written text on internet in the form of mails, tweets etc. consists of shorthand notations. The proposed model maps the popular internet shorthand notations to the corresponding words, thereby making a much more accurate classification as compared to traditional approaches. The proposed model is tested on real benchmark data set and performance is evaluated. The empirical results reveal that the proposed SVM, although is computationally heavier, nevertheless provides an improvement in classification accuracy as compared to those based only on textual features.

Machine Learning, Spam Filtering, Support Vector Machines, Character Level Features