I was struck today by three pieces on predictive analytics
First there was a piece in Computerworld "QuickStudy: Predictive Analytics". This was a pretty nice introduction to predictive analytics (though not as good as my FAQ on the subject of course). There was a really nice summary sentence, however:
Ultimately, businesses want predictive analytics to suggest how to best target resources for maximum return.
Absolutely right - its all about making the best possible use of your resources. I did have one criticism of the article. You could follow all 5 of their best practices (see page 2) and still fail to get any value. The missing best practice is deployment. So read the article, but add this item to the best practices list:
6. Ensure deployment of the models
The best model in the world will not help if you never get it into production. Use technology and processes that focus not just on building a good model but also on making sure the model can be and then is deployed.
Then I saw Oracle's announcement about Sigma Dynamics and Gartner's comment about it - Deal Enables Oracle to Pursue Wider 'Decisioning' Opportunity in which Gareth Herschel says he feels this moves Oracle into "decision management" and then goes on to explain that Sigma Dynamics takes what I would call an Enterprise Decision Management approach by combing rules and analytics, albeit in a specific area. Oracle's website says that
Sigma Dynamics' Real-time Decision software combines customer insight and business requirements to instantaneously make the best recommendation in each customer interaction and operational decision by intelligently adapting to continuously changing information.
Sounds like a localized implementation of an EDM approach to me. Glad to see Oracle joining the party.
And finally, of course, Fair Isaac announced a new release of Model Builder. Version 3.5 now offers Data Spiders™ - genetic algorithm-based technology to help analysts overcome the tedium of data preparation by searching raw transaction data, generating characteristics and finding the most powerful predictors. 3.5 also adds the ability to read and write DBMS and SAS data as well as enhanced support for Predictive Model Markup Language (PMML).
Predictive analytics, like business rules, seems all over the news these days.