Anomaly Detection in Workflow Management

Businesses are constantly seeking new solutions to ensure operational efficiency and maintain orderly processes. Anomaly detection identifies unwanted and unusual situations within an organization’s processes, such as fraud, errors, inefficiencies, and malfunctions. Such anomalies can lead to reduced efficiency and decreased profitability potential for the organization.

As a solution to these issues, Next4biz aims to enhance its BPM product with anomaly detection capabilities. The development of these features is intended to leverage state-of-the-art deep learning, artificial intelligence, and data-driven approaches to identify abnormal conditions within the business processes of organizations across various industries.

By detecting anomalies in their business processes, organizations can increase operational efficiency, prevent potential fraud, and enhance profitability.

One of the studies we conducted as part of this development is the anomaly detection study in business processes titled “An Application on Anomaly Detection in Workflow Management,” published at the II Bilsel Gordion Congress.

In this study, using the Isolation Forest algorithm, our researchers implemented anomaly detection among triplet states, which can be considered sub-processes of business workflows. The Isolation Forest algorithm was run with three different contamination values: 0.01, 0.05, and 0.1.

The analysis was carried out on 46,038 workflow steps recorded in the database of a single operation running on the Next4biz BPM software, with the focus on detecting anomalies in the transitions between workflow steps. Isolation Forest, a low-cost and time-efficient algorithm, was chosen for this purpose, and the outputs obtained at the different contamination values, each representing a different expected proportion of anomalous records, were compared against one another. The findings show that the model successfully identifies anomalies, and the analysis contributes to process optimization by helping organizations make their workflows more efficient.

Looking ahead, our efforts will focus on adapting state-of-the-art anomaly detection approaches, particularly graph-based anomaly detection techniques, to work more comprehensively on business processes. We aim to make these multi-perspective anomaly detection approaches suitable for the Next4biz BPM platform and its tenants, offering users cutting-edge anomaly detection technologies.

Click here for the full text of our study, “An Application on Anomaly Detection in Workflow Management,” published at the II. Gordion Bilsel International Scientific Research Congress.