def predict_proba(self, X):
X1, X2, idx1, idx2 = self._split_data(X)
preds = self.models1[0].predict_proba(X1)
preds[idx2] = self.models2[0].predict_proba(X2.loc[idx2])
for i in range(1, self.num_runs):
preds[idx1] += self.models1[i].predict_proba(X1.loc[idx1])
preds[idx2] += self.models2[i].predict_proba(X2.loc[idx2])
return preds / self.num_runs
def _split_data(self, X):
idx_1 = (X['upd_start_level'] <= START_LEVEL_THRESHOLD).values
idx_2 = (X['upd_start_level'] > START_LEVEL_THRESHOLD).values
X1 = X.loc[:, self.x1_columns]
X2 = X.loc[:, self.x2_columns]
return X1, X2, idx_1, idx_2
if __name__ == '__main__':
np.random.seed(2707)
X_train, X_test, y_train = utils.load_data(data_name='extended')
clf = XGBoostSplit(num_runs=NUM_RUNS, base=CLASSIFIER)
if MODE == 'cv':
utils.perform_cv(X_train, y_train, clf, MODEL_NAME + '-' + CLASSIFIER,
stratify_labels=utils.load_stratify_labels())
elif MODE == 'ensemble':
utils.VJUH(X_train, X_test, y_train, clf, MODEL_NAME, 'ensemble', stratify_labels=utils.load_stratify_labels())
else:
print('Unsupported# mode')
X_train = pd.DataFrame(scaler.transform(X_train), columns=X_train.columns)
X_test = pd.DataFrame(scaler.transform(X_test), columns=X_test.columns)
kf = StratifiedKFold(n_splits=CALIBRATION_CV,
shuffle=True,
random_state=4891)
for i in range(1, MAX_POWER + 1):
clf = KNeighborsClassifier(n_neighbors=2**i, n_jobs=1)
clf = CalibratedClassifierCV(clf, method='isotonic', cv=kf)
utils.VJUH(X_train,
X_test,
y_train,
clf,
MODEL_NAME + str(i),
'ensemble',
stratify_labels=utils.load_stratify_labels())
train = []
test = []
for i in range(1, MAX_POWER + 1):
tr, ts, _ = utils.load_data(data_name=MODEL_NAME + str(i),
data_folder='ensemble')
train.append(tr)
test.append(ts)
result = pd.concat(train, axis=1, ignore_index=True)
utils.save_result(result.values,
MODEL_NAME[:-1],
data_folder='ensemble',
test=False)
result = pd.concat(test, axis=1, ignore_index=True)
utils.save_result(result.values,
import warnings
warnings.filterwarnings("ignore", category=DeprecationWarning)
X_train, X_test, y_train = utils.load_data('log_flipped')
scaler = StandardScaler()
scaler.fit(pd.concat((X_train, X_test), axis=0, ignore_index=True))
X_train = pd.DataFrame(scaler.transform(X_train), columns=X_train.columns)
X_test = pd.DataFrame(scaler.transform(X_test), columns=X_test.columns)
encoder = LabelEncoder()
y_train = encoder.fit_transform(y_train).astype(np.int32)
num_classes = len(encoder.classes_)
X_train = np.array(X_train).astype(np.float32)
X_test = np.array(X_test).astype(np.float32)
num_features = X_train.shape[1]
clf = NeuralNet(input_shape=(None, num_features), output_num_units=num_classes, **PARAMS)
if MODE == 'cv':
utils.perform_cv(X_train, y_train, clf, MODEL_NAME, stratify_labels=utils.load_stratify_labels())
elif MODE == 'ensemble':
utils.VJUH(X_train, X_test, y_train, clf, MODEL_NAME, 'ensemble',
lasagne=True,
runs=RUNS,
stratify_labels=utils.load_stratify_labels()
)
else:
print('Unsupported mode')