Lumnion and HDI Sigora present Cheetah: A Risk Modelling Tool that delivers more accurate results
Risk modelling is at the very heart of insurance. Risk modelling is done generally with tools that use GLM (Generalized Linear Model). Improving the model results is rather difficult, because it is hard to come up with new variables, or combination of variables with the current tools that are available in the market that use GLM. Next to that, there is also a need for audit traceable automated reports for the modelling process, which is being prepared manually at the moment.
But, it is also possible to develop other algorithms and processes using Artificial Intelligence and Machine Learning techniques that yield much more accurate results. Lumnion, together with HDI Sigorta, have developed Cheetah, a Risk Modelling tool for the insurance industry. Cheetah allows sophisticated pricing and risk decisions in a simple and user-friendly environment through its easy-to-use and intuitive user interface. The solution is able to use different dynamic modelling techniques, that include AI and ML algorithms such as GBMs, Random Forest, Decision Trees and various other widely used methods in a transparent way.
The tool can also be integrated into the core insurance systems, allowing it to work online.The state of the art interactive reporting module supports the business in monitoring expected results and strategic analyses. Cheetah matches the sophistication demanded from the end user, and even provides assistance, without ever sacrificing the simplicity in its design. And with this, the tool has already proven its impact at several insurers.
Key benefits of Cheetah
- Cheetah is capable of building conventional GLM models, but still use ML for variable selection.
- It can use pure ML algorithms that offer new, better results in a transparent way.
- It assists the user at every step with innovative advisor systems that provide an added degree of transparency to the process.
- It makes comparison of different ML models easier, allowing for a better informed model selection process.
- It allows for an overall better user experience, regardless of the user’s background.
- It’s fast and easy to deploy in scale.
How it works
One of the key design aspirations that came forward in the early days of Cheetah’s development process was to create a software that has a simple, sleek user interface that is intuitive and easy to use while retaining the functionality and sophistication of modern risk modeling and insurance pricing platforms.
When importing the data set into Cheetah, you can include or exclude certain factors and determine their data type. You can choose the model type, the model column and the GLM type. When provided with a selected subset of features using a variety of machine learning algorithms, Cheetah Advisory will score and rank each selected factor for easier future selection.
Running the model you can see the model results, and you’re also provided with a tab for each model feature allowing further insight of manipulation (see the image underneath).
Machine Learning in Cheetah’s modelling process
ML is where Cheetah provides an edge to the analytical process. Through a modelling tab, in addition to the conventional GLM algorithm that “reigns” in the insurance sector, Cheetah allows the end user to take advantage of more advanced and sophisticated ML algorithms.
Running ML algorithms on data in Cheetah is quite straightforward: start by determining which factors go into the model, select the desired model algorithm and perform hyperparameter optimization to my desire. After this, the model is constructed.
In the model comparison tab, the constructed models are displayed next to each other with their own factsheet – making it easy to compare between different models.
Who is Lumnion?
Lumnion is an Insurtech company that develops AI based pricing tools for the insurance industry. They aim to deliver solutions to our clients for fitting their business for the future, redefine what is possible and find the distinctive ways to move ahead of their competitors.
Check out their DIA Prime Time video here: