يكپارچه سازي و كاوش مجموعه داده هاي حجيم RFID

چكيده انگليسي :

Warehousing and Mining on Massive RFID Data Sets Leila Hafezi l hafezi@ec iut ac ir Date of Submission 26 April 2011 Department of Electerical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language PersianMohammad Hossein Saraee saraee@cc iut ac irAbstractRadio Frequency Identification RFID has been proposed as an efficient effective and useful technology inthe recent years RFID technology is widely used across many application domains with promising resultssuch as supply chain management retail access control airline luggage management medical identification electronic passport and pet identidication It uses radio frequency waves to read a unique identifier that isattached to an unexpensive tag by RFID reader from a far distance and without a line of sight Readerreceives radio waves from tags and convertes them into transformable data and then save data in computerservers This technology facilitates and accelerates many applications but it has proposed a challenge RFIDsystems generates enormous amount of data where traditional methods are not capable of handling them andtherefore new novel and efficient techniques for processing are needed The volume is so enormous thatdisusing the system comes into consideration Data mining techniques are used for modeling of relationshipand discovering hidden pattern in massive data consider being useful RFID data is stored inmultidimensional format by RFID warehousing which is more suitable for analysis and mining Before usingdata mining algorithm we use warehousing techniques to load and store the data in data warehouse In thisproject we have focused on preprocessing techniques and try to improve these techniques for improving datamining In this project a new warehousing model is presented In this new model we add a new step into oldmodel of warehousing which is compression step old model contains data gathering data cleaning and datatransformation and then data warehouse construction This data structure is compressed to the highest degreepossible without missing any data and therefore data warehouse is compressed as well In this method weconsider a definite and deterministic sequence among different phases in the production line of car engines Then we combine these phases at different levels Therefore low level data have been converted to higherand more meaningful levels In this model we save the combination of different levels of stages in the tablein database and thus we can decompose the data at every level and return to level 0 again that is a mainadvantage of our method that is to say no data would be missed in compression This method is applicablein all the systems that do a series of deterministic and repeated operations over data We can get our data aproduct line of car engine to 1 50 times the basic data We can also return to the basic data In addition weare suggesting relinquishing normalization in data warehouse to generate fewer tables within these tables Asshown in thesis this suggestion also improves data mining algorithms execution time We consider 70percent of the data for training and 30 percent for testing and we use 10 fold cross validation for acquiringaccuracy of our method Results show that the error rate is nearly 16 percent We believe it shows goodaccuracy and is acceptable So our method can be used for effective compressing Key wordsRadio Frequency Identification Data Mining Data warehousing and Data warehouse

حافظي، ليلا

يكپارچه سازي و كاوش مجموعه داده هاي حجيم RFID

تكنولوژي شناسايي از طريق امواج راديويي , تكنيك هاي داده كاوي , انبار داده