Nearly every Data Science project is, or should be, formulated the same way. We start with a question, often this question is predictive (What will sales be next quarter?) or explanatory (Why are sales falling?). Predictive problems are easier in the sense that they need only produce a single number as an output, with no requirement to give a human-satisfying rational interpretation. However, predictive problems are tough because you’re inherently asking the machine to interpret a situation it has never seen before (the future) using only data from the past, and because no situation is truly identical to a past situation, the machine is always vulnerable to being blindsided by some totally new phenomenon.
But in either ...
Get How to Talk to Data Scientists now with the O’Reilly learning platform.
O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.
