
Ok,ok…… I’ll admit I really don’t even like fruit. Vegetables,yes. Fruits………… nah.
If you’ve ever spent time with data scientists, (not expecting you’ve ever actually done this), you know they’re full of jargon and technicalities. Heteroskedasticity, confusion matrix, dot product, eigenvectors…. just to name a few.
The same goes for their models. Support vector machines, deep learning, neural networks – yes, these all have their place. The truth is they’re very computationally expensive (ie, they take a LONG time to run) and most have little if any actual interpretations. Can some predictor breast cancer better than a radiologist who’s been practicing for decades? Yes. Can we explain exactly how the model does so much better? Nope.
In business we can’t always just rely on getting a good answer. We need to understand the underlying system and “why things are the way they are.”