In many organisations, there is a need to manage risks (e.g. healthcare). Such systems typically have a large number of risk indicators, and there is a need to trigger an alert when the measures went beyond certain thresholds (e.g. if there is an overcharging of a service or claim). The pain points in such systems include identifying the most relevant or promising risk indicators and setting appropriate risk thresholds. Organisations rely on human experts to do these, which can be biased.
To address this problem, Singapore Data Science Consortium (SDSC) and Institute of Data Science (IDS) jointly worked together with Defence Science and Technology Agency (DSTA) to develop an unsupervised risk management system that effectively and efficiently reduced the need for human intervention in determining thresholds and eliminated potential human bias by applying statistically proven data-driven approach.
Phase 1: Dynamically determine thresholds and eliminate potential human bias
Phase 2: Prioritize and identify new risks
Phase 1 of the project has been completed and DSTA has plans to deploy the system. The team is currently enhancing the system to use advanced machine learning algorithms to automatically identify the most relevant/promising risk indicators, and even detect the unknown unknowns for better risk management. The final system can be deployed to multiple domains like healthcare fraud detection, financial credit card anomaly detection and cyber security network surge activities.
This study received joint contribution from DSTA and IDS.
Project Title: Dynamic Risk Management System
Done By: Defence Science and Technology Agency (DSTA), Institute Of Data Science (IDS), Singapore Data Science Consortium (SDSC)
Contact Person: Dr. Cher Pei Hua, Data Scientist, SDSC
This research is supported by the National Research Foundation, Prime Minister’s Office, Singapore under its Industry Alignment Fund (Pre-positioning) Funding Initiative.