Supervised Semi-definite Embedding for Email Data Cleaning and Visualization

Ning Liu, Fengshan Bai, Jun Yan, Benyu Zhang, Zheng Chen, Wei-Ying Ma

Abstract: The Email systems are playing an important and irreplaceable role in the digital world due to its convenience, efficiency and the rapid growth of World Wide Web (WWW). However, most of the email users nowadays are suffering from the large amounts of irrelevant and noisy emails everyday. Thus algorithms which can clean both the noise features and the irrelevant emails are highly desired. In this paper, we propose a novel Supervised Semi-definite Embedding (SSDE) algorithm to reduce the dimension of email data so as to leave out the noise features of them and visualize these emails in a supervised manner to find the irrelevant ones intuitively. Experiments on a set of received emails of several volunteers during a period of time and some benchmark datasets show the comparable performance of the proposed SSDE algorithm.