Heyvaert, Carl-ErikDarmawan, ViolaStouthuysen, KristofVerdonck, TimGrumiau, ChristopherThoppan Mohanchandralal, Sudaman2023-03-032023-03-032022http://hdl.handle.net/20.500.12127/7193The case is suitable for participants who want to enhance their machine learning (ML) and data science skills by solving a real-world application in the insurance industry. It can be taught to non-executive (i.e., master of business administration) and executive audiences, in courses at the intersection of strategic (management) accounting, data science, and ML, as the case offers a unique example of how data science and ML are applied to the traditional accounting and actuarial field. Although not necessary, it might be useful for participants to have some programming experience in Python before starting the case. After working through the case and assignment questions, students will have the opportunity to do the following: Understand the relevancy of ML within a business context. Identify problematic disability claims and their impact on a company’s P&L. Gain preliminary insights using exploratory data analysis to optimize premium calculation. Apply unsupervised techniques such as clustering and gain insights into the data. Train, use, and interpret supervised ML techniques in a multi-class classification setting. Obtain business-relevant insights to improve smart decision-making and launch targeted strategic actions.During an Allianz Benelux SA (Allianz) board meeting held in early 2019, Allianz’s chief financier officer (CFO) had a profound discussion with Allianz’s chief data and analytics officer (CDAO) on improving the company’s profit and loss (P&L) statement by targeting problematic cases among disability claims related to Allianz’s life insurance product. It appeared that certain claims had very long durations, leading to recurrent payouts surpassing the total amount of premiums. Consequently, there were too many claims that could translate into future losses. If this phenomenon persisted, Allianz could lose millions of dollars in revenues. Therefore, the CFO contacted the CDAO and his data office and requested that the team identify the client segments in which the most problematic cases of disability claims occurred. Additionally, the CFO wanted the data office to build a predictive model that could estimate the duration of a claim, to adapt the premium coverage to specific customer segments.enMachine LearningW27373286358272378119751220053