Survival Data Mining: Predictive Hazard Modeling for Customer History Data

This advanced course identifies the benefits and pitfalls of using survival analysis for business intelligence. Designed for data analysts, it covers both theoretical justification of various survival data mining methods and their practical implementation using SAS software.

Learn how to

• build models for time-dependent outcomes derived from customer event histories
• account for competing risks, time-dependent covariates, censoring, and truncation
• use techniques to model current status data and to evaluate the predictive performance of the model.

Who should attend: Predictive modelers, data analysts, and statisticians

Prerequisites

Before attending this course, you should

• have experience with predictive modeling, particularly with logistic regression
• be familiar with statistical concepts such as random variables, probability distributions, and parameter estimation
• be comfortable working with summation notation, vectors, matrices, and analytic geometry
• have SAS programming proficiency.

Many of the SAS examples use DATA step, macro, and SQL programming. Modeling methods are implemented using SAS/STAT procedures and SAS Enterprise Miner. The Predictive Modeling Using Logistic Regression and Neural Network Modeling courses provide relevant background information. Prior attendance in these courses is advantageous but not required.

Course Contents

Survival Data

• time-dependent outcomes derived from customer event histories
• features of the event-time distribution such as competing risks, time-dependent covariates, censoring, and truncation
• basic nonparametric estimation of the hazard and distribution functions

Flexible Parametric Hazard Models

• multinomial logistic regression for right censored data
• regression spline and neural network modeling
• adaptations for large data sets

Modeling Current Status Data

• simple models for reduced sample data
• powerful models for cross-sectional data

Predictive Performance

• predictive scoring
• estimating the mean residual lifetime
• empirical validation using concentration curves

Software Addressed

This course addresses the following software product(s): SAS/STAT, SAS Enterprise Miner.

This page was created using SAS software.