def fit(self, X, y):
from sklearn.ensemble import RandomForestRegressor
from OptimizedOffsetRegressor import OptimizedOffsetRegressor
self.rfr = RandomForestRegressor(
max_depth=self.max_depth,
n_estimators=self.n_estimators,
n_jobs=self.n_jobs,
random_state=self.random_state,
max_features=self.max_features,
min_samples_leaf=self.min_samples_leaf,
verbose=self.verbose)
self.off = OptimizedOffsetRegressor(n_buckets=self.n_buckets,
initial_offsets=self.initial_offsets,
scoring=self.scoring)
self.rfr.fit(X, y)
tr_y_hat = self.clip(self.rfr.predict(X))
print('Train score is:', -self.scoring(tr_y_hat, y))
self.off.fit(tr_y_hat, y)
print("Offsets:", self.off.offsets_)
return self
class PrudentialRegressor(BaseEstimator, RegressorMixin):
def __init__(self,
max_depth=7,
n_estimators=50,
n_jobs=-1,
random_state=1,
max_features=1.0,
min_samples_leaf=1.0,
verbose=1,
n_buckets=8,
initial_offsets=[-1.5, -2.6, -3.6, -1.2, -0.8, 0.04, 0.7, 3.6],
scoring=NegQWKappaScorer):
self.max_depth = max_depth
self.n_estimators = n_estimators
self.n_jobs = n_jobs
self.random_state = random_state
self.max_features = max_features
self.min_samples_leaf = min_samples_leaf
self.verbose=verbose
self.n_buckets = n_buckets
self.initial_offsets = initial_offsets
self.scoring = scoring
return
def epsilon(self, Tp):
from numpy import finfo
return finfo(Tp).eps
def clip(self, arr):
from numpy import clip
return clip(arr, 0., self.n_buckets * (1. - self.epsilon(arr.dtype)))
def fit(self, X, y):
from sklearn.ensemble import RandomForestRegressor
from OptimizedOffsetRegressor import OptimizedOffsetRegressor
self.rfr = RandomForestRegressor(
max_depth=self.max_depth,
n_estimators=self.n_estimators,
n_jobs=self.n_jobs,
random_state=self.random_state,
max_features=self.max_features,
min_samples_leaf=self.min_samples_leaf,
verbose=self.verbose)
self.off = OptimizedOffsetRegressor(n_buckets=self.n_buckets,
initial_offsets=self.initial_offsets,
scoring=self.scoring)
self.rfr.fit(X, y)
tr_y_hat = self.clip(self.rfr.predict(X))
print('Train score is:', -self.scoring(tr_y_hat, y))
self.off.fit(tr_y_hat, y)
print("Offsets:", self.off.offsets_)
return self
def predict(self, X):
te_y_hat = self.clip(self.rfr.predict(X))
return self.off.predict(te_y_hat)
pass
