Database Reference
In-Depth Information
Similarly, I would say that if you are just developing one model— let's say, a
predictive model—it really is great to try a variety of techniques. You'll learn
a lot. It's like the art and the science. There's the science, and that's actually
the easier part of the whole thing. The art is imagining all the possible signals
you could try as inputs to your model. And trying the different techniques
and seeing what themes arise will get your far. It will help you understand
what classes of signals always seem to pop no matter what technique you
use. Lastly, along with trying different techniques, is to A/B test whenever and
wherever you can, as it's really great to add that to your rigor and understand-
ing of the different techniques and themes.
Gutierrez:
What do you love about data science?
Smallwood:
What I love about it is that it's incredibly creative and innovative.
If someone's just dipping their toes into the field—come on in and learn
more, as it is a fascinating field. If you think, “Oh, it's just going to be some
boring math field,” it's not that at all. It's incredibly challenging and creative.
And that's been a constant surprise to me through the years, and continues to
be. It continues to be more and more creative the longer you're in it, because
you have more tools at your disposal and more intuition, right or wrong, at
different times about the data. And so it just gets more exciting and creative
over time.
