Mining multidimensional contextual outliers from categorical relational data

Abstract

A wide range of methods have been proposed for detecting different types of outliers in full space and subspaces. However, the interpretability of outliers, that is, explaining in what ways and to what extent an object is an outlier, remains a critical open issue. In this paper, we develop a notion of contextual outliers on categorical data. Intuitively, a contextual outlier is a small group of objects that share strong similarity with a significantly larger reference group of objects on some attributes, but deviate dramatically on some other attributes. We develop a detection algorithm, and conduct experiments to evaluate our approach.

Identification and Description of Outliers in the Densmore Collection of Native American Music

“…A recent study of outliers in world music [44] focused on quantitative distributions of outliers in and between countries; outlier descriptions at country level were derived by listening to selected recordings. In this respect the study resembled the situation in wider research on computational outlier analysis which largely neglects outlier description in favour of outlier detection [45].…”

“…Melodic with harmonic framework 1,2,6,10,11,17,20,21,26,27,29,30,32,40,45,57,68,71,74,75,77,79,82,89,91,101,115,121,122,124,143,147,155,165,166 Compared to classical tasks in collection-level music analysis such as clustering [22][23][24][25][26], classification [27][28][29][30][31] and pattern matching and discovery [32][33][34][35][36][37], outlier detection has attracted less attention in music computing research [38]. Outlier detection has been mainly used for cleaning datasets from noisy or erroneous examples in order to improve performance of subsequent classification, music recommendation or sound...…”

Deviation Point Curriculum Learning for Trajectory Outlier Detection in Cooperative Intelligent Transport Systems

“…The standard back-propagation algorithm is used to estimate the input data by group reference function. Tang et al [27] proposed a contextual outlier detection using a similarity-sharing group of points. These points share similarities in some areas and might differ to some others.…”

“…Contextual outlier detection in high and sparse dimensional spaces was performed using parallel computing. Other techniques combine individual outliers into similar clusters [27,29]. Each cluster is then termed as a group of outliers.…”

There and back again: Outlier detection between statistical reasoning and data mining algorithms

“…Many other areas with special requirements could be mentioned, such as categorical or ordinal data (Akoglu, Tong, Vreeken, & Faloutsos, ; Das & Schneider, ; Otey, Ghoting, & Parthasarathy, ; Smets & Vreeken, ; Tang, Pei, Bailey, & Dong, ; Yu, Qian, Lu, & Zhou, ), binary data (Smets & Vreeken, ), or uncertain data (Jiang & Pei, ; Liu, Xiao, Cao, Hao, & Deng, ). Also mining events or trends can be seen as variant of mining anomalies (Schubert, Weiler, & Kriegel, , ).…”