Abstract:
Email transaction logs can be used to track the unusual events which might appear among message conversations. The problem is that to monitor every single email is taking a lot of effort especially when the size of email transaction log is very large. Therefore, it should be a wise option to keep monitoring only the important emails. This research focuses on email message filtering and email content monitoring which are used to create a prototype system of email monitoring in this study. When the filtering percentage and the query terms are obtained, the email message filtering module finds and selects only important messages based on the number of obtained percentage using DFS (Depth First Search algorithm) together with email scoring model. Then, the email content monitoring module analyzes and scores each of email contents comparing to the obtained query terms using VSM (Vector Space Model) together with WordNet database. Finally, the outcome is a list of ranked messages based on their similarity scores. Time to proceed all the system modules rapidly increased when the number of messages was raised up which means that the filtering module can help to relieve the time-consuming problem especially in a large scale of email transaction logs. In addition, users can manually adjust the most suitable filtering percentage provided by the configuration components. The experiments also showed that enhancing terms by WordNet helped in rising up the precision value of the system by increasing the opportunity to match among similar terms. Furthermore, increasing the number of messages to be analyzed also has significantly raised the recall value of the system.