def getModel(self, _params):
return AdaBoost(n_estimators=int(_params['n_estimators']),
learning_rate=_params['learning_rate'],
loss=_params['loss'])
samplenames = np.array(dataframe.columns)
# »ñȡ̽ÕëÃû³Æ
featurenames = np.array(dataframe.index)
# »ñÈ¡±í´ïÖµµÄ¾ØÕó
esetarray = np.array(dataframe)
# ±í´ïÖµµÄ¾ØÕóתÖÃ
esetarray = esetarray.transpose()
# ½«Ñù±¾·ÖÀ࣬
# ·Ö³ÉÈýÀà:PB BM LN
# sampletype = [1]*26+[2]*19+[3]*17
# ·Ö³ÉÁ½Àà:PB BMorLN
sampletype = [0]*10+[1]*10
# Adaboost 方法
# =============================================================================
clf = AdaBoost() # AdaBoost(n_estimators=100)
rfe = RFE(clf, n_features_to_select=1)
# =============================================================================
rfe.fit(esetarray, sampletype)
print("Features sorted by their rank:")
print(sorted(zip(map(lambda x: round(x, 4), rfe.ranking_), featurenames)))
result = sorted(zip(map(lambda x: round(x, 4), rfe.ranking_), featurenames))
resultframe = DataFrame(result)
resultframe.to_csv("ranklist_ABrfe.txt", sep="\t")
if (i == 35):
tmp.append(word)
else:
tmp.append(float(word))
Y.append(tmp.pop())
X.append(tmp[:])
#print("X: ")
#print(X)
#print("Y: ")
#print(Y)
f.close
Z = []
odp = []
for k in range(0, 9):
i = randint(0, 300)
Z.append(X.pop(i))
odp.append(Y.pop(i))
#X = [[0, 2, 4, 6, 8], [1, 3, 5, 7, 9], [7, 9, 11, 13, 15], [6, 8, 10, 12, 14]]
#Y = [1, 0, 0, 1]
#Z = [[8, 9, 10, 14, 15], [4, 5, 7, 11, 13], [4, 6, 8, 10, 12], [3, 5, 7, 9, 11]]
ada = AdaBoost(DecTree(max_depth=1))
ada.fit(X, Y)
print("Predict: ")
print(ada.predict(Z))
print("Probabil: ")
print(ada.predict_proba(Z))
print("Score: ")
print(ada.score(Z, odp))
def getModel(self, _params):
return AdaBoost(n_estimators=int(_params['n_estimators']),
learning_rate=_params['learning_rate'],
algorithm=_params['algorithm'])
if __name__ == '__main__':
# Get raw data
df = pd.read_excel('data/Loan Book Nov-16.xlsx')
X, y, weights = clean_df(df)
# Split data into training Set and Testing Set
indices = np.arange(X.shape[0])
X_train, X_test, y_train, y_test, train_indices, test_indices = train_test_split(
X, y, indices, test_size=0.3, random_state=0, stratify=y)
pipeline = Pipeline([
('rescale', MinMaxScaler()),
# ('model', SGD(loss = 'modified_huber',class_weight = 'balanced', penalty = None)) # classifier
# ('model', RF(class_weight = 'balanced_subsample')) # classifier
('model', AdaBoost(n_estimators=50)) # classifier
])
parameters = {'model__n_estimators': [50, 100, 150]}
best_model = GridSearchCV(
pipeline,
parameters,
cv=5,
verbose=1,
scoring='f1',
fit_params={'model__sample_weight': np.array(weights[train_indices])})
best_model.fit(X_train, y_train)
y_pred_probs = best_model.predict_proba(X_test)
y_pred_probs = pd.DataFrame(y_pred_probs).ix[:, 1]
def __init__(self, C = 1, penalty = 'l2'):
self.C = C
self.penalty = penalty
self._model = AdaBoost(n_estimators = 50)
self._model.set_params(C = C, penalty = penalty)
pass