module Stats: sig .. end
These functions compute several performance measures to compare the
predicted values of a SVM model to the expected values of a
given test data set. For more details, have a look on page 8 in the
LIBSVM paper
val calc_n_correct : Lacaml.D.vec -> Lacaml.D.vec -> int
calc_n_correct expected predicted
Raises Invalid_argument if the vector dimensions do not match.
Returns the number of correctly
predicted labels.
val calc_accuracy : Lacaml.D.vec -> Lacaml.D.vec -> float
calc_accuracy expected predicted
Raises Invalid_argument if the vector dimensions do not match.
Returns the ratio (in percent) of
correctly predicted labels to the number of all labels.
val calc_mse : Lacaml.D.vec -> Lacaml.D.vec -> float
calc_mse expected predicted
Raises Invalid_argument if the vector dimensions do not match.
Returns the mean sum of squared errors.
val calc_scc : Lacaml.D.vec -> Lacaml.D.vec -> float
calc_scc expected predicted
Raises Invalid_argument if the vector dimensions do not match.
Returns the squared correlation coefficient.