Interesting announcements last week from three companies I have reviewed recently – Predixion, Zementis and Lyzasoft. Predixion has developed some interesting cloud-based predictive analytic technology – tools to load data from Excel and build predictive models from it – while Lyzasoft has a nice collaborative business intelligence environment and Zementis has a cloud-based deployment engine for business rules and predictive analytic models (using PMML to bring models in from many modeling tools). The two announcements – Predixion/Lyzasoft and Predixion/Zementis – make an interesting set. With these products you could create a rich decision support/decision management environment in the cloud.
Your data miners and analytic experts could use their favorite tool and produce the output as PMML. You could load that up to the cloud and have it execute there using Zementis. You could load up more data into the cloud, using Predixion and enhance it with the scores calculated from your PMML models before building additional models and doing other sophisticated analytic modeling tasks in a nice Excel-based UI. New models built using Predixion could drive in-the-cloud decisioning (using the Predixion API) when combined with rules using Zementis’ ability to run Drools on the amazon.com compute cloud. And for the transactions you could not handle automatically, your business users could use Lyzasoft to collaborate on figuring out a solution – while retaining access to the same models.
In Smart (Enough) Systems Neil and I wrote about the sliding scale between pure decision support systems (human decision maker supported by a system) and pure decision management systems (automated decision making). These three products can give you a wide range of options long this scale, all in the cloud.
Comments on this entry are closed.
James, you might be right … if only the PAST would predict the FUTURE. As it happens, it doesn’t. So all the predictive stuff is simply nonsense. So automated decision making will be a great way to mess things up in this world even further.
The only way to get things together is to get things back to manageable sizes and improve the real-time transparency for current situations and put it into a context that allows human intuition to make the best of it.
Oh yeah, and certainly we want all our data in the cloud for anyone to mine and make even more rubbush predictions … I am all for it.
Suggested readings are: ‘The Black Swan’, ‘The Drunkard’s Walk’, and ‘The Flaw of Averages’.
Max
Thanks for the comment. I think you fundamentally misunderstand the purpose of predictive analytics and have misread the Black Swan (a book I really like and reviewed here). Predictive analytics are not nonsense and do not attempt to predict the future from averages. This is a common misconception. In fact Predictive analytics turns uncertainty about the future into usable probability. It does not predict the future, it does not eliminate uncertainty or pretend to be definitive. It does give you a way to estimate how likely something is so you can act appropriately. Predictions like this can be used and are useful.
Real-time transparency is clearly useful but human intuition is not a reliable mechanism at any kind of scale. Human intuition is subject to bigotry and bias and has a proven inability to correctly balance recent information with less recent information among other problems. Relying on humans, especially when you would have to rely on large numbers of humans (as most enterprises would) is a much better way to get into trouble.
Banks used to follow your advice in their loan processes. If you were white, male, married and wore a suit you got a loan because the loan officer’s “human intuition” said you were a good credit risk. Personally I prefer a more objective approach. Predictive scores are not perfect (as I discussed here) but they are potentially very useful.
I would refer you to Super Crunchers and check out his meta analysis of the studies that compare people with algorithms – people don’t do well…
JT
The future will never be as the past, but the predictive analises could support some business decisions. Is better to base a decision on facts than on intuition… (well sometimes)
But I really believe on descritive modeling where you try to find hidden knowledge in a huge volume of data!
My 2 cents,
Pedro