Detecting Temporal Anomalies in Business Processes Using Distance-Based Methods

Mavroudopoulos, Ioannis; Gounaris, Anastasios

doi:10.1007/978-3-030-61527-7_40

Cited by 3 publications

(5 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…-As discussed in Rogge-Solti and Kasneci (2014), Mavroudopoulos and Gounaris (2020) it is hard to interpret whether temporal event anomalies are true anomalies or measurement errors. In the latter case, the timestamp of an event has been falsely logged, causing an event to appear as outlier.…”

Section: Discussionmentioning

confidence: 99%

“…Then, this is (most probably) a measurement error and has to be distinguished from real outliers. We have directly tackled this issue in our previous work (Mavroudopoulos and Gounaris 2020), and the main results are orthogonal to the exact proximity-based outlier detection. More specifically, we advocate, after detecting an outlier in a trace, to check whether the immediate successor is an outlier as well.…”

Section: Discussionmentioning

confidence: 99%

“…In this section, we introduce four different variants of outlier detection that can be applied at both the trace and event level. The first three have been introduced in our previous work (Mavroudopoulos and Gounaris 2020), however they were evaluated using only the Euclidean distance. The common feature of all these variants is that, for each trace or event, they perform range queries and examine its neighborhood.…”

Section: Temporal Outlier Detection Based On Proximity Methodsmentioning

confidence: 99%

“…More specifically, we build upon our recent work in Mavroudopoulos and Gounaris (2020), which showed the potential of distance-based outlier detection Ng 1999, 1998) over the statistical approach in Rogge-Solti and Kasneci (2014). Distancebased outlier detection is the most prominent representative of proximity-based methods (Aggarwal 2017) and its main function is the computation of the number of neighbors for each instance.…”

Section: Introductionmentioning

confidence: 99%

“…This definition differs from the one in our previous work inMavroudopoulos and Gounaris (2020), since, in this work we exclusively focus on the temporal attributes of a trace and not on complementary structural aspects.3 If the logs contain the start and end finish time of each task explicitly, then our approach to detecting latency anomalies can be applied to detecting anomalous task durations in a straightforward manner. Also, in that case, the events need not ordered in a strict order.…”

mentioning

confidence: 99%

See 4 more Smart Citations

A comparison of proximity-based methods for detecting temporal anomalies in business processes

Mavroudopoulos

Gounaris

2022

Mach Learn

Self Cite

View full text Add to dashboard Cite

Outlier detection in process mining refers to aspects such as infrequent behavior in relation to the underlying business process models or to anomalous latencies of task execution, termed as temporal anomalies. In this work, we focus on the latter form of anomalies and we aim at investigating in depth the behavior of several proximity-based variants, which are shown to outperform simple statistical ones. We investigate multiple distance functions and approaches to establishing the outlierness of traces or individual tasks, and we explain the superiority of our proposals over existing proximity and probability distribution fitting-based techniques yielding up to 2.05X higher F1 score. We also provide guidelines as to which variant to be chosen based on the type of anomalies targeted and the dataset characteristics.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Temporal Outlier Detection Based On Proximity Methodsmentioning

confidence: 99%