Online Program Home
  My Program

Abstract Details

Activity Number: 425 - SPEED: Reliable Statistical Learning and Data Science
Type: Contributed
Date/Time: Tuesday, August 1, 2017 : 3:05 PM to 3:50 PM
Sponsor: Section on Statistical Learning and Data Science
Abstract #325114
Title: Supervised Binning Techniques for Predictive Modeling
Author(s): Zhen Zhang* and Lei Zhang and James Veillette and Kendell Churchwell
Companies: C Spire and Mississippi State Dept. of Health and C Spire and C Spire
Keywords: target-supervised ; binning ; predictive modeling ; algorithm ; ensemble
Abstract:

For business categories of all kinds, today's advanced technologies and digital processing generates data in an unprecedented speed and volume. This rich data becomes a "gold mine" for strategic marketing teams across industries, as more and more marketers rely on advanced predictive modeling techniques to uncover actionable information for a range of critical marketing decisions. Popular modeling algorithms include decision trees, artificial neural network and traditional statistical modeling methods such as logistic regression models. During our practices of predictive modeling in the telecommunication industry, we find that a certain technique often works better with certain data types better that other, and that an ensemble model of different model types often yield better results than a single model. Moreover, we find that some continuous variables, especially those that do not follow normal distribution, often gains more predicting power when being transformed to an ordinal or nominal type through a target-supervised binning process. In this study we'll introduce the target-supervised binning process and show its improved predicting power through model performance.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2017 program

 
 
Copyright © American Statistical Association