Predictive Analytics Research Reports
Examples of SOA research reports that have made use of predictive analytic techniques.
Health Risk Scoring
- Machine-Learning Methods for Insurance Applications - A Survey
The Society of Actuaries is pleased to make available a research report that provides a literature survey of methodologies applying machine learning to insurance claim modeling.
- Accuracy of Claims-Based Risk Scoring Models
- Risk Scoring in Health Insurance: A Primer
- Uncertainty in Risk Adjustment
- Examining Predictive Modeling–Based Approaches to Characterizing Health Care Fraud
The Health Section Research Committee is pleased to make available a research report providing a systematic evaluation of the modeling methodologies and data samples used to characterize health care fraud. Also included is an Excel file with an inventory of articles and studies reviewed in the report.
- Predictive Modeling of Surface Temperature Extremes over North America, with Actuarial Applications in View
This project obtains ensemble simulations of surface atmospheric temperature over North America. These simulations are used to estimate long-term changes in the spatial distribution and magnitude of extreme heat waves and cold spells in the region.
- Applying Image Recognition to Insurance
The Society of Actuaries announces the release of a new report that will provide a resource for the insurance industry to learn about image recognition technology. Authored by Kailan Shang, the report explores the current status of image recognition technology, models for image recognition, and various possible applications of image recognition to insurance.
- Incorporation of Flood and Other Catastrophe Model Results into Pricing and Underwriting
This research report on the incorporation of flood and other catastrophe model results into pricing and underwriting is prepared by the authors for the Canadian Institute of Actuaries, the Society of Actuaries, and the Casualty Actuarial Society.
- Predicting High-Cost Members in the HCCI Database
Using the Health Care Cost Institute (HCCI) database, which contains claim information on approximately 47 million members annually over a seven-year time period, we examined which characteristics best predict and describe high-cost members. We found that cost history, age, gender and prescription drug coverage are all predictors of future high costs, with cost history being the most predictive.
- Auto Loss Cost Trends
In the latter half of 2013, personal auto insurance carriers began to notice an uptick in property damage liability and collision frequency. This marked the beginning of a new increasing frequency trend bucking over 25 years of falling crash rates. In response, industry partners banded together to analyze these trends. Using publicly available data from the Federal Highway Administration, Bureau of Labor Statistics, the Census Bureau, and other sources, an analysis group is searching for explanatory variables. These materials are an update to the report previously published in January 2018.
- Auto Loss Cost Trends Report
This report provides an analytical basis for discussion and understanding for regulators, legislators, and the general public about national trends and state specific factors that may be driving the auto insurance loss costs, which ultimately impact premiums and the consumer.
- Predictive Models on Conversion Studies for the Level Premium Term Plans
The Reinsurance Section, Product Development Section and the Committee on Life Insurance Research announce the release of a new report summarizing the third phase results of a multi-phase study on term conversions. This report explores conversion rates and post-conversion experience using predictive analytics.