Who Has What? Predictive Modeling Using Customer Billing Data

Table of Contents
Energy Insights Opinion
In This Report
Situation Overview
Applications
Method Specifics
Predictive Modeling Techniques
Logistic Regression
Discriminant Analysis
CHAID
Table: Summary of Predictive Modeling Techniques
Accuracy and Error Rates
Figure: Type I and Type II Errors
General Comments About Modeling
Case Study 1: San Diego Gas & Electric and Central AC
Background and Setup
Logistic Regression Results
Table: Summary of Logistic Regression Models Predicting the Presence of Central AC: SDG&E Case Study
Figure: Predicted Error Rates and Accuracy for Central AC Using Logistic Regression: SDG&E Case Study
Table: Comparison of Logistic Regression Results with Classification Probabilities of 0.5 and 0.4: SDG&E Case Study
Discriminant Analysis Results
Table: Summary of Discriminant Analysis Models Predicting the Presence of Central AC: SDG&E Case Study
CHAID Results
Figure: Partial CHAID Tree for Central AC Model: SDG&E Case Study
Comparison of Methods for Central AC Data
Table: Comparison of Logistic Regression, Discriminant Analysis, and CHAID Models Predicting the Presence of Central AC: SDG&E Case Study
Follow-Up Real-World Application
Case Study 2: Alliant Energy and Electric Heat
Background and Setup
Defining the Criterion Variable
Overall Modeling Approach
Logistic Regression Results
Table: Logistic Regression and Discriminant Analysis Models Predicting the Presence of Electric Heat: Alliant Energy Case Study
Figure: Predicted Error Rates and Accuracy for Electric Heat Using Logistic Regression: Alliant Energy Case Study
Discriminant Analysis Results
CHAID Results
Figure: CHAID Tree for Electric Heat Model: Alliant Energy Case Study
Comparison of Methods for Electric Heat Data
Lessons Learned
Future Outlook
Essential Guidance
Actions to Consider
Learn More
Related Research
Synopsis