Scalable pattern mining and analysis in real-time RIFD data

基于射频识别（RFID）的技术最近已经成为与安全、物流、零售和商业智能相关的几个重要应用的关键组成部分。其中，基于RFID的实时定位系统（RTLS）能够实时识别、定位和跟踪资源，显示出巨大的商业潜力。然而，这些技术也带来了重大的科学和技术挑战，因为大量的数据不断实时产生，超过了现有技术的存储、管理和分析能力。本研究的目标是通过引入有效和可扩展的方法来分析和检测实时RFID数据中的异常模式，提高基于RFID的RTLS系统的商业价值。为了满足本文所针对的大规模、开放和动态环境的可扩展性、可重复性和适应性要求，我们提出了一种基于潜在语义分析和集成学习的新方法。与现有的方法不同，这些方法将能够发现在原始数据流中不能直接观察到的高级和更通用的模式。为了评估我们提出的方法的有效性和效率，我们将在机场安全的高度重要的工业环境中进行测试。为此，我们将与Purelink Technology公司合作，该公司是基于rfid的RTLS系统开发和商业化的全球领导者，他们将为验证我们的方法提供必要的技术，数据和专业知识。该项目对工业界和科学界的预期效益是巨大的。因此，它将导致有效的实时分析方法的发展，为多维RFID数据提供更好的信息。此外，基于rfid的检测异常行为的新技术的发展将有助于提高大规模环境中人口的安全性，例如机场，工厂和发电厂，通过提供低成本，自动和非侵入性的方式来检测潜在的威胁，否则无法被当前系统检测到。

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.