RFID systems should be designed to process a large number of RFID data in real time. Therefore, there have been much research and company studies regarding RFID data processing. One of methods is CEP (Complex Event Processing), which can provide a method to process RFID data efficiently. However, previous work is just focused on raw RFID data processing, such as data filtering, the elimination of duplicated data, and the aggregation of data. Also, it creates primitive events based on just one physical or logical reader. Therefore, processing overhead for complex events may increase. And it cannot provide business level events. Therefore, we propose a method that can reduce processing overhead and create business level events by using CEP. Proposed method provides two primitive events that are defined by using the relationship of two readers. Thus, when any pattern of events is matched for a specific complex event, business level events can be generated and execution time can be reduced comparing with other mechanisms without those events. And, execution time can be reduced by about 57% as compared to others.