Research On Mining Method Of RFID Frequent Path Event Pattern

RFID(Radio Frequency Identification) is a kind of uncontact automatic recognition and data access technology. Now has been widely used in many industries such as smart building, supply chain management, border inspection, theft detection and so on. RFID system will generate a large number of RFID basic events, there are a lot of problems as false negative, false positive and repeated read in basic events, that can't be used in application directly. Although there are many methods to due with the RFID basic events streams cleaning, but no one could reach the 100% cleaning effects. Based on this unthoroughly cleaned events streams that generate the composite events have the non-deterministic feature.Cleaning of composite events is a difficult problem. Through our study that owing to the higher level of the composite events that is difficult to cleaning if no other reference information, only use the primitive data cleaning methods is unable to get the correct results. To find the reference information of the composite events cleaning, we studying many of the RFID application. Found in these applications will generate a large number of timing label data. These data is very important for find labels'Frequent path event model in system. This paper with the idea of data mining make mining frequent path event model that support composite events cleaning from the timing relational data as our research . The work are as follows.(1) In the time window first to implement the get of RFID path events. Includingâ‘ Path identify.Take EPC as the key values, and by the time relevance of the data that has the same EPC to find a single object moving track(a path event). That make the rambling RFID path information into a suitable form to mining .â‘¡P ath information Compression. Greatly save the storage space by compressed the RFID path information from time and location dimensions.(2) Propose a frequent probability path model. The model is the abstract of the frequent path event pattern that have timing relationship in RFID application. In this paper use sliding window to dynamicly update the frequent path event model. The frequent path model can not only reliably predict the follow-up events but also can support the cleaning of the complex event that have timing relationship.(3) This paper propose a method of mining the frequent closed path patterns in RFID data that SW-MFCP(Sliding Window-Mining Frequent Closed Pattens), this Algorithm use the time dacay mechanisms to dynamicly mining the frequent path model within the silding window, With the forward of time window,use the PM-tree(Path Mining Tree) to capture the latest model information in the event streams.(4) To describe the possibility of moving of the object in the path branch, this paper concern on each node in the path event model, compute Transition probabilities and condition emission probabilities,that Introduce the time constraints into the model, and give the occurring probabilities for a node that is a termination of a path event. Last discussed the frequent path event model's importance to composite events that have timing relations .(5) This paper verify the probability frequent model's effectiveness by experiment. and verify the SW-MFCP algorithm that based on the time dacay mechanism is effectiveness in time and space.