PREDICTIVE MODELING IN FINANCE & INSURAN

This course introduces to the students, generalized linear models (GLM), time series models, and some popular statistical learning models such as decision trees models as well as random forests and boosting trees.  The aim for GLM is to provide a flexible framework for the analysis and model building using the likelihood techniques for almost any data type. The aim for the statistical learning models is to build and predict or understand data structure (if unsupervised) using statistical learning methods such as tree-based for supervised learning and the Principle Component Analysis and Clustering for unsupervised learning.  It develops a student’s knowledge of the theoretical basis in predictive modeling, computational implementation of the models and their application in finance and insurance. Tools such as cross-validation and techniques such as regularization and dimension reduction for fitting and selecting models are explored.  We also implement these models using a combination of Excel and R.

The class covers the material of Exams, Statistics for Risk Modeling (SRM) and Predictive Analytics (PA) of Society of Actuaries, and some material of Exams, Modern Actuarial Statistics I (MAS-I) and MAS II by the Casualty Actuarial Society. This is a core course for the Actuarial Science students. Students who have already taken and passed the SRM and PA exams administered by the SOA are exempted from this class and can substitute an elective.