How should I design my data warehouse in order to develop successful data mining applications?
Your best bet is to keep things simple. Unfortunately most data mining software applications have a very unsophisticated view of the data that they mine. Most would like to see a single table, with a single record per "entity" (patient, customer, offer, etc.). This is probably not practical for realistic data warehouse implementations.
One alternative is to find a data mining system that will work with more complex data (one data point is the Oracle data mining capabilities, which have started moving in this direction). The other alternative is to stage the data in such a way as to allow you to build data tables in a format the mining tool will work with, using ETL (Extract, Transform, and Load) tools or scripting languages such as perl.
Dig Deeper on Topics Archive
Related Q&A from Kurt Thearling
Have a question for an expert?
Please add a title for your question
Get answers from a TechTarget expert on whatever's puzzling you.