CODEWITHSUNDEEP
machine-learningscikit-learn
Elimination
Elimination

Reputation: 2723

Find out the training error after fit()

I'm training a LinearSVC model and I want to get the training error of it. Is it possible to get it w/o evaluating it manually?

Thanks

Upvotes: 0

Views: 929

Answers (1)

sascha
sascha

Reputation: 33532

sklearn is using liblinear for this task.

You can take a quick glance into the sources here:

self.coef_, self.intercept_, self.n_iter_ = _fit_liblinear(
    X, y, self.C, self.fit_intercept, self.intercept_scaling,
    self.class_weight, self.penalty, self.dual, self.verbose,
    self.max_iter, self.tol, self.random_state, self.multi_class,
    self.loss, sample_weight=sample_weight)

which shows that only coefficients, intercepts and number of iterations are processed by sklearn's python-API. Whatever else is available in liblinear's output is not grabbed. You can't directly read out the training-error without changing the internal code.

There might be a possible hack turning on verbose-mode, redirect the output and parse additional info available there. But this assumes the info you look for is available there and it's also hacky and i won't recommend it.

Just use the score-method. It won't be too costly compared to fitting.

Upvotes: 1

Related Questions