METHOD AND APPARATUS FOR MINING REAL-TIME FREQUENT EVENT ON STREAM DATA

摘要

An apparatus for mining a real-time frequent event according to the present invention comprises: a conversion module for converting a generated event into a predefined event attribute, when an event is generated from inputted stream data, and storing the event in a window of a predetermined size; a candidate generation management module for combining all pattern candidates which can be generated and are currently stored by a tree with an inputted event, when an event is inputted from the conversion module, so as to store and manage all frequent pattern candidates which can be generated in a current window, in a candidate tree; and a frequent event extraction management module for receiving a candidate pattern stored in the candidate tree of the candidate generation management module and reflecting the received candidate pattern to the frequent pattern tree to store the same. According to the present invention, since an instant event generated at a moment from stream data and a persistent event continued for a long period of time are expressed as data and used for frequent pattern mining, a relationship between two types of data can be found. In addition, a frequent pattern can be found in real-time by processing the stream data at an inputted moment without dividing the stream data in a batch unit.;COPYRIGHT KIPO 2019