Model Tuning¶

This part of the package provides a simple API for model-tuning routines.

gen_cross_validate(evalfun, n, model)¶

Perform in parallel a generic cross-validation (CV) routine defined in evalfun by the splitting specified in model.cv_gen.

Parameters:	evalfun – function to evaluate n – total number of data points (instances) to create `Kfold` CV generator if `model.cv_gen` is undefined (null) model – `SALSAModel` which contains the `cv_gen` field of type `Nullable{CrossValGenerator}` [1] or `model.output.cv_folds` field containing predefined indices for each fold
Returns:	an average of `evalfun` evaluations.

misclass(y, yhat)¶: Calculate misclassification rate as $\frac{1}{n}\sum_{i=1}^n I(y_i \neq \hat{y}_i)$ .

mse(y, yhat)¶: Calculate mean squared error as $\frac{1}{n}\|y - \hat{y}\|^2$

auc(y, yhat[, n=100])¶: Calculate Area Under ROC Curve. Default number of thresholds is 100.

Footnotes

[1]	wrapper around the type defined in MLBase.jl package