Abstract
Data science is increasingly adopted in actuarial research and insurance practice, yet its reliability depends on every stage of the workflow, from data preparation to model building to operational decisions, in a regulated, data-sensitive industry. This talk follows that pipeline through three connected contributions. We begin with data preparation, showing that conventional practices like random train-test partitioning produce unstable results on highly imbalanced insurance loss data. We propose a statistically principled alternative: support points for representative splitting, Chatterjee's correlation coefficient for non-parametric feature screening, and integrated missing-data handling. Evaluations on simulated and benchmark datasets demonstrate improved robustness and interpretability at lower computational cost. We then embed these methods in InsurAutoML, an insurance-tailored automated machine learning framework whose data balancing, ensemble pipelines, and customized loss functions let non-expert users deploy robust models with a few lines of code. Finally, moving from prediction to decision support, we use millions of historical warranty claims to fine-tune locally deployed large language models via Low-Rank Adaptation, generating structured corrective-action recommendations from unstructured claim narratives. Domain-adaptive fine-tuning substantially outperforms general-purpose LLMs, with about 98% of cases closely matching ground truth. Together, these results point toward trustworthy, governable automation across the insurance analytics pipeline.
About the Speaker
Zhiyu (Frank) Quan is an Assistant Professor at the Department of Actuarial and Risk Management Sciences of the University of Illinois Urbana-Champaign, a Brad and Karen Smith Professorial Scholar of the College of Liberal Arts and Sciences, and a Finance and Insurance Sector Lead for Discovery Partners Institute. He holds a Ph.D. in Actuarial Science from the University of Connecticut. Before joining Illinois, he worked in a cutting-edge Insurtech company as a R \& D data scientist developing data-driven solutions for major insurance companies. He has a broad spectrum of research interests in data science applications in insurance such as tree-based models, natural language processing, GenAI, and applies his actuarial expertise to build predictive models for claim management, rate making, etc. His research agenda is driven by real-life data and is inspired by collaborations with InsurTech, regulators, and insurers. Besides, he is a director of the Illinois Risk Lab, which facilitates research activities that integrate academic training with practical problem-solving in real business settings. He has received 2021 Arnold O. Beckman Research Award, 2025 Bob Alting von Geusau Prize, 2025 NAAJ Annual Prize for Best Paper, 2026 CCIRM Best Paper Award, and has been awarded by the Society of Actuaries Research Institute and the Casualty Actuarial Society.