Description: A review of the Bridging the Gap Series: Application of Predictive Modeling in VA/FIA Risk Management at the Equity Based Guarantees Conference in Chicago.

Description: This article describes a method that can be useful when re-fitting a risk score model to a population of only moderate size. Learn how to use concepts like ridge regression to train a model with partial credibility.

Description: The well-known examples of predictive modeling “successes” are numerous. Perhaps less widely recognized is the myriad of ways in which a predictive model can fail to perform as expected, often due to misconceptions or misrepresentations on the part of the analyst. This article, which is intended for a general actuarial audience, focuses on three such pitfalls. By understanding these three common pitfalls, readers may guard against them within their own work, teams, and organizations.

Description: This actually is a copy of an appendix of a research paper sponsored by the SOA Committee on Finance Research. You can read the entire research paper: “Lapse Modeling for the Post-Level Period—a Practical Application of Predictive Modeling” at the SOA site https://www.soa.org/Research/Research-Projects/Finance-Investment/lapse-2015-modeling-post-level/#sthash.W9lERSls.dpbs, but you can read this valuable appendix copied here in this issue to see how to build a PM step-by-step using your data and the R programming language. We thank Richard Xu, Dihui Lai, Minyu Cao, Scott Rushing, and Tim Rozar for their excellent paper and the Society of Actuaries, for the permission to reprint this portion of the paper.

Description: Spreadsheets have become ubiquitous in business and forecasting. However, they do not usually have the same level of control that any other complex program would require. As a result, we often get erroneous results which can have far reaching, negative consequences.

Description: An introduction of predictive modeling techniques that are applicable to insurance industry, including generalized linear model (GLM) and classification and regression tree model(CART). Other models such as data clustering, neural network, and support vector machine (SVM), are also briefly discussed.

Description: Discusses how 'insignificant' changes in assumptions can have very significant affects.

Description: A discussion on how to use data visualization for model controls.

Description: Even the best algorithms are only effective if they are trained on appropriate data. This article discusses common pitfalls in creating a training data set.

Description: This article goes through some basic data exploration techniques to show how different approaches look.