An intelligent RFID data predicting method based on BP-adaboost

Combinatorial Radio Frequency Identification (RFID) benchmarking test methodology is proposed to instruct users to deploy RFID real-life systems by giving reliable benchmark data. However, it is very inefficient, since collecting RFID data is usually very time-consuming. With the rapid increase of RFID components, combinatorial RFID benchmarking tests get more and more expensive. To address this problem, an intelligent RFID data predicting method is proposed. The predicting model for each reader antenna in the RFID system is learnt from a small set of real RFID data by the BP-Adaboost algorithm. The model then is used to produce highly accurate RFID reading results at different antenna deployments. The RFID system performance can be estimated from these reading results, which can be used to help users choose the suitable RFID system for certain application. Conducted experiments showed that such predicting model produced RFID data with an accuracy over 92%, and can be used to quickly find out improper RFID components and systems for different applications in a very short time.