By Topic

A Graph-Based Approach for Multi-folder Email Classification

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$31 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

3 Author(s)
Chakravarthy, S. ; Dept. Of Comp. Sci. & Eng., Univ. of Texas at Arlington, Arlington, TX, USA ; Venkatachalam, A. ; Telang, A.

This paper presents a novel framework for multi-folder email classification using graph mining as the underlying technique. Although several techniques exist (e.g., SVM, TF-IDF, n-gram) for addressing this problem in a delimited context, they heavily rely on extracting high-frequency keywords, thus ignoring the inherent structural aspects of an email (or document in general) which can play a critical role in classification. Some of the models (e.g., n-gram) consider only the words without taking into consideration where in the structure these words appear together. This paper presents a supervised learning model that leverages graph mining techniques for multi-folder email classification. A ranking formula is presented for ordering the representative - common and recurring - substructures generated from pre-classified emails. These ranked representative substructures are then used for categorizing incoming emails. This approach is based on a global ranking model that incorporates several relevant parameters for email classification and overcomes numerous problems faced by extant approaches used for multi-folder classification. A number of parameters which influence the generation of representative substructures are analyzed, reexamined, and adapted to multiple folders. The effect of graph representations has been analyzed. The effectiveness of the proposed approach has been validated experimentally.

Published in:

Data Mining (ICDM), 2010 IEEE 10th International Conference on

Date of Conference:

13-17 Dec. 2010