| Summary: | 碩士 === 輔仁大學 === 資訊管理學系 === 94 === Enterprise Resource Planning systems, which integrate key resources of the enterprise, become one of the most important information systems for enterprises. But the amount of data in ERP systems is numerous and jumbled, which causes long download times (this is one of the reasons why the users resist to use ERP systems). If we can download reports that users need in advance, it will not only improve the efficiency but also increase the user satisfaction when users utilize ERP systems.
Especially in the WWW environment nowadays, we have to avoid information postponement. Users do not like to wait for a long time for downloading when they retrieve information on the Internet. Therefore, in our research, we adopt data mining techniques to construct a mechanism for analyzing, learning and predicting users’ behavior automatically. The mechanism we proposed predicts users’ next click and download in advance reports that users might need. This reduces users’ waiting time for downloading reports and increases the ease to use of the system.
Before conducting our experiment, we interviewed experts in enterprises for further understanding user’s report clicking behavior of ERP systems. In our experiment we simulated users’ behavior based on the results of the interview. Experiment result shows when user behavior is serialized, the precision is up to 76%. Therefore, the enterprise can decide when to use our prefetch system according to the serial characteristics of the report clicking and the regularity of the time fields in users’ behavior. In addition, the way of analyzing user behavior in our research could be a start point of developing personalized ERP systems in the future.
|
|---|
