Prestigious Machine Learning Body Invites Report Authors to Present Insurance Reserve Accuracy and Prediction Paper

January 2022

DOCOsoft, working with academics from Dublin’s Technological University (TU) and CeADAR Technology Centre for Applied AI, Dublin, has published a review of insurance reserve prediction techniques used in economics and actuarial science literature as well as machine learning and computer science literature.

The paper, entitled Insurance Reserve Prediction: Opportunities and Challenges, was accepted at the international conference on Computational Science and Computational Intelligence (CSCI), following peer review from leading machine learning academics. This organisation is part of the American Council on Science and Education. The paper was presented at the December 2021 hybrid conference, held in Las Vegas and virtually.

The report concludes that machine learning-based methods, which rely on nonlinear predictive techniques, are a huge opportunity for insurers and could significantly outperform existing stochastic actuarial methods if the data can be integrated into insurers’ technology platforms.

In the report, the authors say: “Predicting claims’ reserves is a critical challenge for insurers and has dramatic consequences on their managerial, financial, and underwriting decisions. The insurers’ capital and their underwriting capacity of further business are subject to the unexpected reserve estimation. Increasing premium rates and adjusting the underwriting policy decisions may balance the impact of unexpected claims. Consequently, this will implicate their business opportunities negatively.”

The authors state that: “more data could enhance the understanding and representation of prediction models; thus, they improve the prediction accuracy.”

The report concludes with several potential directions for future travel in this exciting and emerging field of insurance Machine Learning. The authors say that enhancing the data pre-processing should improve the estimation process to increase data quality. Improving models promises to improve the processing and accuracy of reserve prediction methods. Other machine learning prediction approaches are on the horizon which include incorporating unstructured data such as insurance reports, social behaviour data, spatio-temporal data, and image & multimedia data.

DOCOsoft CEO Aidan O’Neill commented: “This is a very prestigious honour for DOCOsoft and demonstrates the value of academic collaboration with TU Dublin as well as support from Enterprise Ireland with the Marie Curie Fellowship. It also highlights the potential this academic collaboration offers our employees in terms of academic achievement and professional development. We are increasing our investment in insurance claims R&D and this recognition of our thought leadership output, in a crowded field of insurance technology innovation, is very welcome.”

The report’s authors are, as follows:

Ayman Taha Technological University Dublin, Dublin, Ireland and Cairo University, Egypt

Bernard Cosgrave DOCOsoft Dublin, Ireland

Wael Rashwan Technological University Dublin and CeADAR Technology Centre for Applied AI Dublin, Ireland

Susan McKeever Technological University Dublin and CeADAR Technology Centre for Applied AI Dublin, Ireland