CODEWITHSUNDEEP
machine-learningbinaryclassification
Daniel
Daniel

Reputation: 2500

Binary classification, one or two targets?

I was wondering if there is a more or less correct way to set your targets when doing binary classification. Let's say I have a time series and I want to predict if the next entry will be UP or DOWN.

Personally I would lean towards setting this up with just one target, [1,0]; 1 for UP and 0 for DOWN. But I've also seen examples with two targets, [1,0] for UP and [0,1] for DOWN. Which (if any) is the correct approach?

Upvotes: 0

Views: 78

Answers (1)

lejlot
lejlot

Reputation: 66805

You should use one target, as this is the simplest model that expresses all problem's constraints (Okham's razor). Two outputs will work too, but there is nothing to be gained, while you increase (by tiny factor, but you do) your memory consumption, computational complexity etc.

Upvotes: 2

Related Questions