About 3 years ago, I wrote about the Petabyte Age, relating to an article in Wired.
In the latest Wired UK edition, Aug-12, I read The Exabyte Revolution, Neal Pollak.
About the Age of Big Data. Business now view data as raw material -- an economic input a par with capital and labour.
But as we know, they are worthless, unless they can be interpreted. We need to extract knowledge, or meaning from them. We still talk about machine-larning.
I remembered, Hal Varian, one of the highlighted actors in the article, compiled papers into a book, 1992, Economic and Financial Modeling with Mathematica.
It described how the Black-Scholes models can be implemented in Mathematica with a background on stochastic calculus (Mathematica and Diffusion) and how the Black-Scholes PDE is derived through Ito's Lemma ...
He is now chief economist at google specializing in information economics. He introduced "Correlate" a relatively new feature from Google - one of the statements in the article really sounds "home-made" to me: to get better forecasting from predictive modeling you need to run a whole bunch of predictive models in parallel - applying multi-strategy and multi-method approaches with a lot of cross-model validation.
This was our objective from the beginning of mlf. And we have achieved this before the petabyte age emerged and the new computing muscles with massive parallelism will help us to apply the approach to really massive data.
Parallel model generation and cross-model validation is of good nature for inherent parallelism.
Time to call mlf a platform for data-science? Approaching data -scientists - the new rock stars in information technologies?
Coming back to Black-Scholes option modeling. To calculate fair prices of exotic options, more complex models are applied introducing stochastic volatility and it is a challenging task to recalibrate such models, especially calculation stochastic implied volatility - on the fly - to market data.
An inverse problem that combines modeling and data-driven methods.
When this approach becomes common sense massive market data should be provided to the community. Derivatives and risk analytics will then be able to provide better decision support (including systemic risk assessment).
But at the moment the financial modelers and the data-scientists do not seem to see the potential of shared approaches and scientific co-evolution?
