Time series clustering via matrix profile and community detection

Unsupervised constrained discord detection in IoT-based online crane monitoring

“…In contrast, unsupervised approaches do not require a labeled dataset for anomaly identification. Some of the common approaches include the use of adversarial networks [32], variational transformers [33], Graph Convolutional Adversarial Networks [34], fuzzy c-means and Markov models [35], Gaussian Mixture Models [36], matrix profile and community detection [37], and others [27,23,38].…”

“…Similarly, Ilonen et al [27] use discriminative energy functions to identify failure-prone frequency-domain regions to enable a generic motor condition diagnosis tool. Li et al [37] use matrix profiles and social network techniques, such as community detection for time series clustering. Wang et al [33] capture long-term dependencies on telemetry time series data and perform anomaly detection on them using variational autoencoders and transformers.…”

Process bottlenecks identification and its root cause analysis using fusion-based clustering and knowledge graph

“…Traditional community detection approaches are always based on graph features, such as nodes, edges and paths [30]. The combination of graph embedding and clustering provides a new approach to community detection [31,32].…”

Early faint fault diagnosis of wheelset axlebox bearings in urban rail trains based on ICiSSA-MOMEDA

“…It gives K-means lower computation time and higher clustering accuracy for multidimensional AR models by using a fast EM iterative algorithm. The k-ARs are performed in the following steps; the detailed calculation process is described in the literature [35,36].…”