We are delighted to showcase the “Lux War Risk Model”. The model is a multi-layer perceptron. This is a supervised feed-forward neural network with multiple layers which is trained to predict whether the country is in the conflict status or not, based on several input features sourced from reliable platforms such as the World Bank, International Labour Organisation, UN Population Program and the REIGN (Rulers, Elections and Irregular Governance) Database among others.
It may be useful to create early warning predictors for armed conflicts and potentially use them to understand what drives conflicts and wars. Such indicators may play a role in risk pricing.
This is an excellent opportunity to learn how models may help to make sense of an uncertain future.
Session details:
When: Thursday 4th May 2023 1000 – 1130 GST
Where: Online
Registration: Please email us info@difcia.org
Regards,
Simon S. Price
CEO DIFC Insurance Association
Synopsis
Civil war and organised violence continue to erupt throughout the world into the 21st century, with thousands of victims annually. In addition to the direct damage inflicted such conflicts also have a negative impact on the socio-economic environment, including but not limited to economic growth, willingness to do business and demographics. Armed conflict appears to be more likely in some countries than in others, and we use a data-driven approach to understand this likelihood and its main drivers. Retrospectively, for each conflict, the root cause or a complex of such causes can be assessed. It therefore can be useful to create early warning predictors for armed conflicts and potentially use them to understand what drives conflicts and wars. Such early warning indicators can play a key role in P&C ratemaking, especially in the Marine, Construction and Property LOBs.
Our model is designed not only to determine the degree of risk of conflict for each country in the near future but also to show the main objectively measurable indicators that can be linked to the emergence of conflicts that have already occurred and may affect the development of conflict in the future. The key engine of the Model is a Multi-layer Perceptron. This is a Supervised Feed-forward Neural Network with multiple layers which is trained to predict whether the country is in the conflict status or not, based on several input features sourced from reliable platforms such as the World Bank, International Labour Organisation, UN Population Program and the REIGN (Rulers, Elections and Irregular Governance) Database among others.
The key production output from the Model is a world map where countries are ranked in accordance with the risk profile assessed by the model based on the likelihood of the conflict in each month in the next 3 years (up to the end of 2024):
- Low Risk: Likelihood of conflict does not exceed 25%
- Low to Moderate Risk: Likelihood of conflict is between 25% and 50%
- Moderate to High Risk: Likelihood of conflict is between 50% and 75%
- High Risk: Likelihood of conflict exceeds 75%.
The Model outputs for each country the assigned likelihood of being in each state (in conflict / not in conflict) which allows explicit ranking.
Guest Speakers
Ruan van Rensburg
Founder
Ruan is the founder of Lux Actuaries and supporter of innovative ideas and new developments at Lux. He also concerns himself with the growth of Lux into new countries especially in the developing world, and establishing Lux as the premier actuarial consultancy in the countries we operate in.
Navarun Jain
Senior Actuarial Analyst
Navarun Jain is a Senior Analyst and Data Science specialist at Lux UAE. As Vice-Chair of the CAS Machine Learning Task Force, he is part of a global team contributing to definitive research on Machine Learning and advanced data analytics for the actuarial profession. Navarun is an AWS Certified Machine Learning Specialist and lead developer of the Lux CWS Model.