How Advanced Statistical Methods and Data Analytics Help Understand Impact of Major Risk Factors to Insurance Losses

Visualization and interpretability issues have been an ongoing challenge, particularly in complex statistical data analysis. In these two papers, I studied how to apply advanced statistical techniques to capture the pattern of insurance loss data distribution. I investigated relative claim frequency and Size-of-loss functionality using industry-level loss data. The study of the Size-of-Loss distribution is significant because it serves as a benchmark for insurance and reinsurance companies. The estimation of distributional behaviour has a huge impact on the auto insurance industry.
On the other hand, within the modelling with generalized linear models, a new variable importance measure for the categorical risk factors is proposed to quantify their variable importance. To overcome technical difficulties caused by traditional approaches, I developed new variable importance measures for factors included in the generalized linear model. This makes our modelling approach more comprehensive for application to real-world predictive modelling due to this capability in capturing variable importance.
The proposed method presented in this research helps improve data explainability and understanding of the overall auto insurance Size-of-Loss distribution pattern. On the other hand, our proposed method of variable importance measures shows great success in ranking variables in terms of importance. It coincides with other state-of-the-art methods. Furthermore, they are applicable to similar data analysis in many business applications, including economics and finance.
The impact of this research is to help develop strategies to control and manage the proposed rate changes by insurance companies for rate regulation purposes. In addition, the research provides better insights for audiences regarding the nature of insurance pricing and a better understanding of the fairness of auto insurance premiums.
Our findings on variable importance measures of major risk factors can help insurance regulators to better understand the impact of risk factors on the model response when modelling loss amounts or claim losses. In addition, this research contributes to further developing statistical modelling techniques to make them even more practical. The results obtained from this research provide benchmark estimates to facilitate insurance rate regulation.To learn more, see the full articles:
Shengkun Xie (2022). Feature extraction of auto insurance size of loss data using functional principal component analysis. Expert Systems with Applications, Volume 198. DOI: 10.1016/j.eswa.2022.116780 (external link)
Shengkun Xie and Rebecca Luo. (2022). Measuring Variable Importance in Generalized Linear Models for Modeling Size of Loss Distributions. Mathematics 10, no. 10: 1630. DOI: 10.3390/math10101630 (external link)