String labels for classification

Question

I've recently started playing SciKit ML. I just got my hands on classification algorithms (SGDClassifier, LinearSVC) and I'm not sure how to properly represent feature labels.

Suppose I'm trying to predict a football match outcome. Given N features the result is either 1=WIN or X=DRAW or 2=LOSS. Can/should I label my features using string values?

Eg.:

X_train = [[n0, n1, …, nn], …]
y_train = ['1', 'X', '2', …]

Or maybe I should use boolean matrices instead, like [win,draw,loss], eg:

y_train = [[1,0,0], [0,1,0], [0,0,1], …]

Or maybe I should use a single digit to map each label 1=win and 0=draw, -1=loss, eg:

y_train = [1, 0, -1]

Which one would be best to pick and why? I'm a newbie, thank you for your patience.

Answer 1

Mutually exclusive classes are usually integers; in your case, win=0, draw=1, loss=2 (for instance).

It is mostly a convention and is convenient for implementation.