A survey on outlier detection in Internet of Things big data
Abdullah A. Al-khatib, Mohammed Balfaqih, Abdelmajid Khelil · Institution of Engineering and Technology eBooks · 2019
In this chapter, more than one criteria are combined for better review and characterization of OD techniques. First, we classified the solutions based on big data phase criteria into data-generation and data-acquisition phases. Then, we categorized the techniques that detect outlier in data-acquisition phase on outlier type into fault detection, event detection, and intrusion detection. We classified the fault-detection techniques based on OD method into statistical based, machine learning, distance based, and density based. Thereafter, based on if the techniques assume the underlying distribution model and estimate the parameters of the model or not, are classified into parametric based and nonparametric based, respectively. We classified the machine learning techniques depending on if the user influences machine-learning technique or not, into supervised and unsupervised techniques, which are classified further based on the analyzing approach. Moreover, we categorized the distance-based and density-based techniques based on distance and density measurements.