Volume 16 How To Detect And Handle Outliers 22.pdf

Wang et al. [86] presented a PCA, as well as separable compression sensing, to identify different matrices. Compressive sensing (or compressed sampling [CSG]) theory was proposed by Candes and Wakin [87], that uses a random measurement matrix to convert a high-dimensional signal to low-dimensional signal until the signal is compressible after which the original signal is restructured from the data of the low-dimensional signal. Moreover, the low-dimensional setting contains the main features of the high-dimensional signal, which means CSG can provide an effective method for anomaly detection in high-dimensional data sets. In the model of Wang et al. [86], abnormalities are more noticeable in a matrix of uncompression compared to a matrix of compression. Hence, their model could attain equal performance in volume anomaly detection, as it used the original uncompressed data and minimized the computational cost significantly.

Their model was compared with the ZMad, Med100 and C&K150 methods using real and synthetic data sets, as well as a subjective evaluation by the pilot. Wang et al. [86] reported that the ZMad method outperforms all other methods for outlier detection in real data sets, as they reported that it attains better detection rates and has the potential to reduce the time to complete analysis. The experimental results also illustrated that ZMad has the highest detection rate at a lower false detection rate than the Med100 and C&K methods.

Whereas previous studies have pointed out the facial features of outliers and the possible causes, we aimed to verify these possible causes. Eysenbein et al. [104] investigated the topological properties of the data, i.e., the node and link densities, of the network looking at Flickr members’ relationships. Similarly, Steinhaeuser et al. [105] were interested in the topological properties of the transportation networks in the United States to examine vulnerabilities and failures. The authors reported that outliers were primarily on the network periphery, where the density of links and nodes was less than in the core part of the graph. d2c66b5586