Browsing by Author "Lu, Tennyson"

Now showing 1 - 2 of 2

A Study to Improving Anomaly Detection in a TS Workflow
(2016)
Wang, Feng-Jian
;
Thanh-Thuy Thi Nguyen
;
Lu, Tennyson
;
資訊工程學系
;
Department of Computer Science
The analysis for the behavior of the artifacts adopted in a workflow can help preventing the execution errors. A temporal (structured) workflow, TS Workflow, is described with the min and max execution time intervals for each process and modeled by control structures (Sequence, AND, XOR, Loop). In the past, the anomalous behaviors of artifacts in a workflow are redefined to simplify artifact analysis in a workflow. However, some anomalies detected in the previous work might not occur if the temporal factors are considered in each process, because their corresponding processes cannot run continuously or concurrently due to temporal factors. In this work, the analysis work is improved by filtering out above anomalies in a TS workflow. Our work also handles the loop structure that complicates the analysis work and was not concerned in conventional analysis works.
Improving Workflow Anomaly Detection with a C-Tree
(2017-01-01)
Wang, Feng-Jian
;
Chang, Alex
;
Lu, Tennyson
;
資訊工程學系
;
Department of Computer Science
To guarantee the correctness of workflow execution, it is essential to analyze the structural and artifact integrity of workflows. The current best approach of artifact workflow anomaly detection is O(vertical bar E vertical bar) for structured workflows, however, each of the anomalies returned in the approach contains (artifact, operator) at each workflow node. In this paper, we present an innovative methodology which contains the following two characteristics: 1) A C-Tree (defined in Section 3) structure which separates sequential and parallel issues in workflow analysis and increases the convenience and elegancy of anomaly detection; and 2) A loop-reduction method which helps lower the size of nodes to be analyzed while not losing the abilities of detecting anomalies within workflow models. The anomaly detection is done by 1) transforming the BPMN into the C-Tree, 2) and detecting the anomaly in the C-tree. Compared with current best approach, 1) Our method can show the first operator and its location of an anomaly detected directly, although it cannot speed up the execution time, 2) The execution times of anomaly detection inside loop is decreased, and 3) Our method can detect concurrent (parallel) workflow anomaly based on C-Tree.