CAS Monograph No. 6: A Machine-Learning Approach to Parameter Estimation
A Machine-Learning Approach to Parameter Estimation, the sixth volume of the CAS Monograph Series, is now available for download. In this monograph, CAS Fellows Jim Kunce and Som Chatterjee address the use of machine-learning techniques to solve insurance problems. Their model can use any regression-based machine-learning algorithm to analyze the nonlinear relationships between the parameters of statistical distributions and features that relate to a specific problem. Unlike traditional stratification and segmentation, the authors' machine-learning approach to parameter estimation (MLAPE) learns the underlying parameter groups from the data and uses validation to ensure appropriate predictive power.
The authors present an implementation of this model for the lognormal distribution utilizing the K-nearest neighbors, kernel regression, and relevance vector machine algorithms, which incorporate the concepts of training/testing/validation data sets, parameter sweeps, outlier removal, and Bayesian maximum a posteriori estimation. They then demonstrate the ability of the machine to learn different clusters of mu and sigma from a publicly available closed claim data set. With an understanding of this machine, actuaries will be prepared to incorporate machine-learning algorithms into their actuarial work product and compete in the era of "big data."
Jim Kunce is Senior Vice President and Chief Actuary at MedPro Group. He holds B.S. degrees in Physics and Astronomy from the University of Kansas and a M.S. in Applied Computer Science from Purdue University. He is a Fellow of the Casualty Actuarial Society and a Member of the American Academy of Actuaries. Jim began his actuarial career in 1994 and has also held positions at Kemper Insurance and GE Insurance Solutions. While at GE Insurance, he earned his Master Black Belt certification in Six Sigma Quality and was issued a patent for methods and structure for improved interactive statistical analysis.
Som Chatterjee is President and CEO of 121 Mapping Inc., a Delaware company focused on building analytics launch-pads and mobilizing data assets to better predict outcomes and prescribe actions for insurance carriers and intermediaries globally. He holds a bachelor's in statistics from the Indian Statistical Institute. He is a Fellow of the Casualty Actuarial Society and a Member of the American Academy of Actuaries. Som began his actuarial career in 2004 and has held positions at Genpact and MedPro Group. Som can be contacted via email at firstname.lastname@example.org, or visit www.121mapping.com.
CAS monographs are authoritative, peer-reviewed, in-depth works focusing on important topics within property and casualty actuarial practice. The CAS Monograph Series aims to create important additions to the existing body of CAS literature, with each monograph treating a single subject comprehensively. The Monograph Editorial Board (MEB) manages the monograph publication process in close coordination with the CAS editorial staff. Submission guidelines can be found on the CAS website or by contacting Donna Royston at email@example.com.
This monograph is currently available to all CAS members as a free download. Print copies are available in the CAS online store.
The CAS Monograph Series showcases CAS members' extensive specialized expertise, helping to raise the performance standard for property and casualty actuaries through insightful research. The monographs represent just one way that the CAS provides its members with access to relevant information, research and resources that they can apply directly on the job to advance in their careers. For information on submitting a monograph, visit the Monograph Submission Guidelines page.