We recently had the pleasure to be the main organizer of the Bucharest Edition of the Global AI Bootcamp on the topic: Detecting Insurance Claims Fraud with Machine Learning. The Bootcamp is a worldwide initiative taking place in over 50 cities simultaneously.
During the event, our experts partnered with Microsoft to deliver practical workshops and hands-on applications tailored to the Insurance and Banking industries.
Fraud classification and detection is a key endeavor for financial services companies in their search for an optimal and timely manner to manage risks. For example, AI can be applied when coupled with a Custom Vision scenario, in which the claim details, such as the parts of the goods, which have been damaged, the severity and other information, are automatically determined by such a model and fed into a fraud detection model afterwards.
As a centrepiece of the event, we featured a practical workshop on Detecting Insurance Claims Fraud with Machine Learning | Softelligence. This data experiment was developed by Softelligence Data Enablement Consultant, Laurentiu Diaconu. Laurentiu has played a significant role in many of our data projects, including IFRS implementations, and is a leading member of our Machine Learning Practice.
Technology: Microsoft Azure Machine Learning Studio
Microsoft Azure Machine Learning Studio is a collaborative, drag-and-drop tool you can use to build, test, and deploy predictive analytics solutions on your data. The Studio publishes models as web services that can easily be consumed by custom apps or BI tools such as Excel. Machine Learning Studio is where data science, predictive analytics, cloud resources, and your organization’s data meet.
Machine learning is an application of artificial intelligence (AI) that blends algorithms with statistics to find patterns in huge amounts of data. Any type of data – numbers, images, clicks, which can be digitally stored – can be fuel a machine learning algorithm. You can experience machine learning applications in many of the services that we use today: recommendation systems, search engines, social media feeds, voice assistants and the list goes on.
In all of these instances, each platform is collecting as much data about the customers as possible and uses machine learning to make accurate decisions or predictions, for example about what customers might want next. Machine learning algorithms bring strengths such as the ability to cut through complexity that is different from but at the same time complementary to, human skills.
Results of our Detecting Insurance Claims Fraud with Machine Learning model
The model developed during this interactive application exercise was able to detect fraud at various probability levels, based on the data used in the training session. New claims can be submitted to the service for classification. The app uses the information returned from calls to the Web Service to tell if the claims are fraudulent or not. It also shows the probability that the classification assigned to each claim is correct. It is important to note that the more claims are used to train the model, the more accurate its probability rate will be.
Fraud classification and detection is a key endeavor to be tackled by Financial Services companies in their search for an optimal and timely manner to manage risks. This use-case can best be applied when coupled with a Custom Vision scenario in which the claim details such as parts of the goods which have been damaged, severity and other information, are automatically determined by such a model and fed into a Fraud Detection model afterwards.
However, as companies own such siloed data already, this model can be rapidly trained and made available to production systems in an expedited manner.