def transform(self, X, y=None):
if X.shape[1] == 0:
return X
encoded = pd.get_dummies(todf(X), dummy_na=True,
drop_first=False,
sparse=self.sparse)
logger.debug('dummy encoded shape {}'.format(encoded.shape))
return align_columns(encoded, self.encoded.columns)
def _check_params(self, X):
p = X.shape[1]
self.k_int = ratio2int(p, self.k)
self.donothing = (self.k_int <= 0 or self.k_int >= p)
if self.donothing:
self.feature_names = todf(X).columns
else:
self.feature_names = ['svd' + str(i) for i in range(self.k_int)]
def fit(self, X, y=None):
if X.shape[1] == 0:
self.encoded = X
return self
self.encoded = pd.get_dummies(todf(X), dummy_na=True,
drop_first=self.drop_first,
sparse=self.sparse)
logger.debug('dummy encoded shape {}'.format(self.encoded.shape))
return self
def fit_transform(self, X, y=None):
self._check_params(X)
if self.donothing:
return X
self.svd = TruncatedSVD(n_components=self.k_int)
return todf(self.svd.fit_transform(X))
def transform(self, X):
if self.donothing:
return X
return todf(self.svd.transform(X))