Big data is profoundly changing the lifestyles of people around the world in an unprecedented way. Driven by the requirements of applications across many industries, research on big data has been growing. Methods to manage and analyze big data to extract valuable information are the key of big data research. Starting from the variety challenge of big data, this dissertation proposes a universal big data management and analysis framework based on metric space. In this framework, the Hilbert Index-based Outlier Detection (HIOD) algorithm is proposed. HIOD can handle all datatypes that can be abstracted to metric space and achieve higher detection speed. Experimental results indicate that HIOD can effectively overcome the variety challenge of big data and achieves a 2.02 speed up over iORCA on average and, in certain cases, up to 5.57. The distance calculation times are reduced by 47.57% on average and up to 89.10%.
A Novel Framework for Context-aware Outlier Detection in Big Data Streams
