Data Clean And Cluster Analysis For Path Data From RFID Agricultural Products Tracing System

RFID technology is widely used in supply chain systems, manufacturing industry, transportation management with the features of scanning fast, small, diverse forms, penetration and good safety. Items with RFID tags will generate a lot of RFID movement path data during the movement, extracting useful information and knowledge becomes an important research topic.This paper describes the RFID technology and its principles, and analyzes the characteristics of RFID data path, builds and designs RFID technology in all aspects of agricultural products tracing system combining RFID agricultural products tracing systems. However, due to radio frequency identification technology is sensitive to the external environment, resulting in lower accuracy of reading rate and uncertainties of data. You need to clean the RFID data before storing data.Uncertainty data cleaning algorithm DSMURF is proposed which is based on dynamic RFID tags. We build a real experiment platform to analyze the relationship between the reading rate and the tag distance, antenna angle, tag speed. The adaptive sliding window algorithm SMURF set too large windows for dynamic tags. The result will cause huge errors. DSMURF algorithm sets the size of the sliding window by tag speed, reader epoch, the reader’s read range. And the other, SMURF has no redundant data processing. We propose a framework for cleaning RFID redundant data. Finally, we verify the effectiveness of the algorithm DSMURF by build real experiment platform and simulated data sets.In this paper, we propose RFID data stream clustering algorithm RCluStream based on online micro-clustering and offline macro-clustering. The measure of path similarity is the basis of cluster analysis. We give the definition of RFID path similarity. RFID path data have the features of stream, the clusters constantly change because of data flows during the process of clustering. Online micro clustering set up cluster feature vectors, if the cluster is changed, then recording cluster feature vectors; Meanwhile, offline macro clustering return to the cluster feature vectors and changes of cluster according to the user input parameters.This paper develops RFID agricultural products tracing system interface and client interface. The system visualizes the path data and expresses the path distribution and data mining results paths. Experiments demonstrate the effectiveness of the RCluStream algorithm.