WHITE PAPERS AT A GLANCE
Classing/Binning of Variables in Scorecards
Resource Center
Looking for a comprehensive source of analytic information that can help turn your business challenges into opportunities? We have what you need, from white papers to industry information.
If you don't find what you're looking for, please contact us.
White Papers
Classing/Binning of Variables in Scorecards
One distinguishing characteristic of scorecards is their use of binned predictors. While several approaches to binning predictors are supported by Xeno®, this paper details a data-driven procedure for binning interval-scaled predictors that utilizes an outcome variable to optimize the bin boundaries.
top >>>Scorecard Development in Xeno
This paper provides an overview of scorecard development and discusses how this process is handled in Xeno. A range of topics — including multiple performance objectives (multiple outcome optimization), classing, stability, and palatability — are discussed.
top >>>Performance Inference
The need for performance inference arises in cases where a valid performance measurement is not available for some fraction of the sample that represents the population of interest for a predictive model. This paper focuses on the practice of performance inference and best practices for its application through Xeno.
top >>>Collinearity in Modeling
Consumer credit and retail marketing data are correlated by nature. Statistical models developed on correlated data often exhibit collinearity. It is well-known that collinearity can impact the ability of a model to explain the relationships between the predictor variables and the outcome variable. This paper discusses the differences between explanatory and predictive modeling, reviews the theory of collinearity in the context of predictive modeling, and describes the effects and non-effects of collinearity on scorecard models developed in Xeno.
top >>>Use of Trees for Modeling and Strategy Design
Due to their unique visual structure, tree models and strategies have been widely used in many industries. This document provides a detailed description of tree models and their common uses, and describes their unique implementation in Xeno.
top >>>Clustering in Xeno
Clustering is the partitioning of observations into subsets or clusters, such that the observations in each cluster are similar in profile to each other based on a set of user-specified variables. While different technologies exist for conducting clustering, Xeno employs a variation of the well established k-means clustering algorithm which overcomes many of the limitations associated with the traditional k-means methodology. Xeno enables incorporation of discrete variables, accounting for missing and special values, and handling of variable scaling, outliers and multicollinearity. This paper provides an overview and background on the unique algorithmic and diagnostic features of clustering in Xeno.
top >>>