A Prediction Mechanism of Mail Retrieval Based on the User Behavior Analysis for Electronic Mail Database System

Due to the frequent information exchange, electronic mail system is not only a mail exchange platform but also becomes an information center. Users usually save their mails permanently in mailboxes and retrieve them as need. In the database strategy, each mail is separated into several components and stored in the database individually. Whenever a user queries a specific mail, the mail server retrieves all the components from the database and make up a mail. By using database strategy, it is easy to develop new applications such as search and mobile retrieval. However, the recombination process is time-consuming and causes the response time between query and replay is long. In order to reduce the response time of mail retrieves, we proposed a prediction mechanism built in the mail server. The prediction mechanism predicts the mails that users would access with a high possibility. The mail server reconstructs these mails and saves them in cache. The prediction is based on the user behavior collection and analysis. In our implementation experience, we find that the response time would be reduced and the system performance does not reduce in evidence.