Assisting the appraisal of e-mail records with automatic classification

This spreadsheet shows the results of training a series of Support Vector Machine (SVM) classifiers to replicate the decision process for identifying e-mails of business value from among 846 e-mails in two corporate mailboxes. One set of SVMs were trained on emails without including the text of attachments and with the inclusion of attachments.  For the sake of comparison, these SVMs were also trained on Spam / Ham and on a randomly selected subset of Enron emails half of which were arbitrarily tagged as "business value".