def fit_local(self, X, Y=None):
"""Fitting and generating the local space.
Parameters
----------
X : matrix of shape = [n_samples, n_features]
(i.e., the feature matrix)
Y : matrix of shape = [n_samples, n_outputs]
(i.e., the label/output matrix)
"""
if self.method == 'rf':
local = RandomForestRegressor(n_estimators=self.n_est,max_features='sqrt',max_depth=None, min_samples_leaf=self.stop_crit,random_state=0)
print("Basic model: Random Forest \n")
else:
local = ExtraTreesRegressor(n_estimators=self.n_est,max_features='sqrt',max_depth=None, min_samples_leaf=self.stop_crit,random_state=0)
print("Basic model: Extremely Randomized Trees \n")
if Y is None:
local.fit(X,X)
print("Unsupervised learning \n")
else:
local.fit(X,Y)
print("Supervised learning \n")
treepath = local.decision_path(X)[0]
w = treepath.sum(0)
wlog = np.log(w.astype(float))+0.00001
local.cw = np.power(wlog,-1)
treepath = treepath.multiply(local.cw).toarray().astype(float)
# treepath = treepath.toarray().astype(float)
local.ind = np.where(w<(X.shape[0]*self.dw))[1]
# treepath = np.delete(treepath,local.ind,axis=1)
treepath = treepath[:,local.ind]
local.pca = PCA(self.dim)
local.treepath = local.pca.fit_transform(treepath)
return local
