Risk Detector -testing advanced analytics to support audits
Tölli, Mari (2022)
2022
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2022120426261
https://urn.fi/URN:NBN:fi:amk-2022120426261
Tiivistelmä
This thesis investigates the feasibility of using a software-based data-analysis tool in auditing processes. The possible benefit of using the software in audit processes is that it could help auditors to find possibly interesting anomalies and relationships from large data. This could help with workload and improve the quality of audits.
Risk Detector is a tool that combines advanced analytics and visualization software. Risk Detector was created in a hackathon and has not been changed since then. Therefore, some technical solutions have become outdated, and some changes need to be made to the tool’s architecture to move it under the organization’s control. The first goal of this project is to change Risk Detector so, the organization can gain full control of the tool and have a working version of Risk Detector for testing. The second part of the project focuses on testing the potential of Risk Detector and finding out, how it could be used in the organization.
Bipartite network embedding (BiNE) tries to recreate the original network as well as possible, so it recreates both direct and indirect relationships of the network. Direct relationships can be found from the observed edges. Indirect relationships cannot be directly observed but they are found based on the unobserved but transitive links. Vertex embeddings are calculated using a joint optimization to combine both objective functions of direct and indirect relationships.
Isolation Forest is an algorithm that detects anomalies in data. It is based on decision trees and its anomaly detection is built based on the fact that anomalies are few and different compared to normal instances. Risk Detector uses the BiNE and the Isolation Forest algorithm in its data analysis.
In the new architecture of Risk Detector Azure cloud storage was replaced with on-premises SQL server. In some operations SSIS replaces Python code. Data analyses of the tool are still based on the BiNE and the Isolation Forest algorithm, but some changes were made to Python code to adapt it to changes in the architecture. Microsoft Power BI replaced Tableau as a visualization software. All visualizations were recreated as like the old one as possible.
Risk Detector was tested by the group of the organization’s employees. The main goals of the testing were to find potential use cases for Risk Detector and report any problems found from the tool. The test group reported some issues they found from Risk Detector. The test group also suggested 10 potential use cases for Risk Detector.
Risk Detector is a tool that combines advanced analytics and visualization software. Risk Detector was created in a hackathon and has not been changed since then. Therefore, some technical solutions have become outdated, and some changes need to be made to the tool’s architecture to move it under the organization’s control. The first goal of this project is to change Risk Detector so, the organization can gain full control of the tool and have a working version of Risk Detector for testing. The second part of the project focuses on testing the potential of Risk Detector and finding out, how it could be used in the organization.
Bipartite network embedding (BiNE) tries to recreate the original network as well as possible, so it recreates both direct and indirect relationships of the network. Direct relationships can be found from the observed edges. Indirect relationships cannot be directly observed but they are found based on the unobserved but transitive links. Vertex embeddings are calculated using a joint optimization to combine both objective functions of direct and indirect relationships.
Isolation Forest is an algorithm that detects anomalies in data. It is based on decision trees and its anomaly detection is built based on the fact that anomalies are few and different compared to normal instances. Risk Detector uses the BiNE and the Isolation Forest algorithm in its data analysis.
In the new architecture of Risk Detector Azure cloud storage was replaced with on-premises SQL server. In some operations SSIS replaces Python code. Data analyses of the tool are still based on the BiNE and the Isolation Forest algorithm, but some changes were made to Python code to adapt it to changes in the architecture. Microsoft Power BI replaced Tableau as a visualization software. All visualizations were recreated as like the old one as possible.
Risk Detector was tested by the group of the organization’s employees. The main goals of the testing were to find potential use cases for Risk Detector and report any problems found from the tool. The test group reported some issues they found from Risk Detector. The test group also suggested 10 potential use cases for Risk Detector.