Introduction to Survival Data Mining
Do you want to learn about survival data mining?
This webinar (see attached slides as well) will discuss topics such as building models for time-dependent outcomes derived from customer event histories, accounting for competing risks and incorporating time-dependent covariates.
Demonstrated in SAS/STAT® 14.3, you’ll see how to fit a multinomial logistic regression model with splines for right-censored data and how to produce graphics that illustrate results of a model.
The goal of survival data mining is to model the distribution of time until an event. In the above graph, the lines are the customers and there are two types of events. The horizontal axis is the calendar date and the vertical axis is the length of time the customer is in the database at the date of the event. Note that different customers can have the same length of time in the database when the events occur even when the dates of occurrence of the events are different. Therefore, it is the length of time in the database (otherwise known as tenure or event time) that matters as we are modeling in absolute time.
The precise definition of the outcome event depends on the industry, the specific business practices of the company, the available data, and the intended interventions or remedial actions. The logical inception date is the earliest date that the outcome can occur. The results of the analysis are expressed in terms of a chosen time scale. Consequently, they should represent the smallest units that are measurable and actionable. Typically, the values represent days, weeks, months, billing cycles, quarters, or years.
