/* ---- Google Analytics Code Below */

Friday, October 25, 2013

Big Data Explored and Vindicated

Review of:   Big Data: A Revolution That Will Transform How We Live, Work, and Think by Viktor Mayer-Schonberger, and Kenneth Cukier.   Strongly recommended, for reading through or browsing.

This books does an excellent job of making the Big Data argument.  That we have changed from a fundamental world of Why?  to a  world of What?  And that by simply having more data we can, in many cases, depend on correlation that does not require causality.

Key sections:

Now:  Why are we at this point now, what is driving us to the use of more data productively using analytics.  Some excellent examples. Letting the Data Speak.
More: Why is More Data Better?  Why should we rely on using some of the data when the sample size can be all of the data?  How this facilitates drill down into the data.
Messy: More data introduces more messiness.  We no longer have to remove the messy data, but can learn to live with them as part of the landscape.    Inexact is often a correct measure.   More can trump better when you seek experiment.
Correlation: Still does not imply causation.  But correlation can provide a meaningful way to relate data and create patterns.  Statisticians may cringe, but we are not seeking physical truth here, but useful relationships to solve real problems.  
Datafication: We can look to means of gathering more data, more completely and more often to drive understanding.   This way we can also track dynamic systems that are constantly changing.
Value: Data has real value.  It can be logged, related to other data, licensed and sold.   Much data is also being made public free, to be made useful by visualization and other analytical methods. The value can be in the data: Private, public when it is mashed together with analytic methods.

The Implications, Risks and Future of Big Data are further explored. Lots of great examples of all of these. The best general, non technical book I have read on the subject.

No comments: