Scalable pattern mining and analysis in real-time RIFD data

点击翻译按钮获取中文摘要

Technologies based on radio frequency identification (RFID) have recently become a key component of several important applications related to safety, logistics, retail, and business intelligence. Among these technologies, real-time location systems (RTLS) based on RFID, which enable to identify, locate and track resources in real-time, show a great commercial potential. However, such technologies also bring significant scientific and technical challenges, due to the large amount of data continuously generated in real-time, exceeding the storage, management and analysis capacities of current technologies. The goal of this research is to improve the commercial value of RFID-based RTLS systems by introducing efficient and scalable methods to analyze and detect abnormal patterns in real-time RFID data. In order to meet the scalability, repeatability and adaptability requirements of the large-scale, open and dynamic environments targeted by this proposal, we propose to develop novel methods based on latent semantic analysis and ensemble learning. Unlike existing approaches, these methods will enable the discovery of high-level and more general patterns that are not directly observable in the raw data streams. To evaluate the usefulness and efficiency of our proposed methods, we will test them in the highly important industrial setting of airport security. For this purpose, we will work in collaboration with the company Purelink Technology, a global leader in the development and commercialization of RFID-based RTLS systems, who will provide the technology, data and expertise essential for the validation of our methods. The expected benefits of this project to the industry and scientific community are numerous. Thus, it will lead to the development of efficient real-time analysis methods that provide better information on multidimensional RFID data. Also, the development of new RFID-based techniques to detect abnormal behaviors will contribute to increase the safety of the population in large-scale environments, such as airports, factories and power plants, by providing a lower cost, automatic, and non-intrusive way of detecting potential threats, otherwise undetected by current systems.