Adopting Predictive Models New to Property-Casualty Insurance

January 19, 2017 Insurance Topics 0
image_pdfimage_print
Actuaries are adopting predictive models new to property-casualty insurance

Actuaries Are Experimenting with Predictive Models New to Property-Casualty Insurance

Actuaries are finding predictive models new to property-casualty insurance and discovering new applications for traditional analytics such as Generalized Linear Models (GLMs).

According to my recently published Actuarial Review article, Predictive Modeling: Actuaries Blaze New Analytical Frontiers, these fresh analytical approaches offer great potential for improving the efficiency and profitability for property-casualty insurance companies.

The article is part II of my series on the latest in predictive modeling. The piece describes how GLMs and decision trees are seeing a greater use for applications including claims triage and stochastic loss reserving. Part I, The Quest for Data Gold, covers how actuaries are taking advantage of greater data availability.

By reading the article, you’ll also learn how decision trees are also growing in sophistication. Actuaries are applying those for finding patterns, anomalies or errors for data exploration. To improve the claims process, decision trees helpful for automating subrogation potential discovery, detecting fraud and optimizing report ordering for underwriting.

Predictive models new to property-casualty insurance include unsupervised models and machine learning analytics such as gradient-boosting methods (GBMs), genetic algorithms and random forests. Such models are already in use for fine-tuning pricing, improving market segmentation, determining the likelihood of writing and retaining clients, discovering competitive position and anticipating expected market profitability.

In most cases, choosing which type of model is optimal for a particular application remains in the experimental phase. There are hundreds of models to explore. Both predictive models new to property-casualty insurance and traditional ones offer advantages and disadvantages. All of these require careful consideration.

_______________
…choosing which type of model is optimal
for a particular application
remains in the experimental phase. 

_______________

Predictive modeling tends to advance in personal lines such as personal auto before making their way to commercial lines. However, actuaries using models for commercial insurance coverage are finding opportunities for lines ranging from business owners policies (BOP) to workers’ compensation.

While the potential of predictive models new to property-casualty insurance, traditional GLMs and decision trees is exciting, there are implementation impediments. Part III of my series in the latest of predictive modeling will describe those barriers, include a conversation on data ethics and explain the future data and analytics driven insurance business model. The article will be published in the March/April 2017 issue of Actuarial Review.

What advancements in predictive analytics do you find exciting? Please leave a comment below.