-
Book Overview & Buying
-
Table Of Contents
-
Feedback & Rating
Data Analysis Using SQL and Excel - Second Edition
By :
The first modeling technique is look-alike models, which are used to measure similarity to a known good or bad instance.
The look-alike model produces a similarity score. The model itself is a formula that describes the similarity, and this formula can be applied to new data. Typically, the purpose of a look-alike model is to choose some groups of customers or zip codes for further analysis or for a marketing effort.
The similarity measure cannot really be validated quantitatively. However, we can qualitatively evaluate the model by seeing if the rankings look reasonable and if the historical ranking of something we care about—such as response—is similar to the rankings provided by the similarity score.
This example starts with the question: Which zip codes have the highest penetration of orders and what are some of their demographic characteristics? For practical purposes, the zip codes are limited to those...
Change the font size
Change margin width
Change background colour