def __init__(self, **params):

self.clf = None

# self.num_boost_round = params['num_round']

self.params = params

self.params.update({'objective': 'binary:logistic'})

self.params.update({'silent':1})

def fit(self, X, y):

num_boost_round = self.params['num_round']

# num_boost_round = 1900

dtrain = xgb.DMatrix(X, label=y,missing=-1)

# params = self.params

if X.shape[1]==1:

self.params.update({'colsample_bytree': 1.0})

self.clf = xgb.train(

self.params,

dtrain,

num_boost_round=num_boost_round

)

self.fscore = self.clf.get_fscore()

# bb=np.zeros(dtrain.num_col())

bb = {}

for ftemp, vtemp in self.fscore.items():

# bb[int(ftemp[1:])]=vtemp

bb[ftemp] = vtemp

# bb=bb/float(bb.max())

i = 0

cc = np.zeros(dtrain.num_col())

for feature, value in bb.items():

cc[i] = value

i+=1

self.coef_= cc

def predict(self, X):

# dX = xgb.DMatrix(X)

m,n = X.shape

x_labels = np.array([2]*m)

dX = xgb.DMatrix(X, x_labels, missing=-1)

y = self.clf.predict(dX)

return y

def score(self, X, y):

Y = self.predict(X)

return self.rmse_loss(y, Y)

def scoreAUC(self, X, y):

Y = self.predict(X)

return self.auc_score(y,Y)

def get_params(self, deep=True):

return self.params

def set_params(self, **params):

# if 'num_boost_round' in params:

# self.num_boost_round = params.pop('num_boost_round')

if 'objective' in params:

del params['objective']

self.params.update(params)

return self

def rmse_loss(self,y,y_pred):

return rmse(y,y_pred)

def auc_score(self, y, y_pred):

sorted_y_pred = sorted(y_pred,reverse=True)

return auc(y,sorted_y_pred)