Data mining approaches to modelling insurance risk
Inna Kolyshkina, Richard Brookes · 2002
1 Introduction 3 2 What is data mining? 3 3 Why use data mining? 4 4 Data mining methodologies – Decision trees (CART), MARS and hybrid models 5 5 Case study – Claims streaming in Workers Compensation insurance7 6 Case study – Predicting hospital cost in health insurance 12 7 Appendix - Brief description of CART 16 8 References 18 Table 1 Misclassification tables for the learning and test data 9 Figure 1 Gains chart for the CART model 10 Figure 2 Claim Streaming Evaluation 11 Figure 3 The gains chart for overall annual hospital cost 14 Figure 4 The bar chart of averaged actual and predicted values for overall annual hospital cost 15 Figure 5 Scatter graph 16