Mining Multilevel Association Rules on RFID Data

In SCM, the problem with RFID data is that the volume increases according to time and location, thus, resulting in an enormous degree of data duplication. Therefore it is difficult to extract useful knowledge hidden in data using existing association rule mining techniques, or analyze data using statistical techniques or queries. However, strong associations discovered at high concept levels may represent common sense knowledge and RFID data represented as a concept hierarchy has the property that the data size at the lowest level increases in proportion to the item group. This paper has two aims. Firstly, we use time generalization to eliminate data duplication. Generalization is useful in data mining since they permit the discovery of knowledge at different levels of abstraction, such as multilevel association rules. Secondly, to reduce the complexity of rule generation by examining association rules limited to the level of interest of the consumer, not all concept hierarchy level on a each concept level have its own level passage threshold. As a result, rule generation time is reduced and the query speed is significantly accelerated, due to filtering of data.