def test_microgbt_train_predict():
num_iters = 100
early_stopping_rounds = 10
X_train, X_valid, y_train, y_valid = load_titanic()
# Train
gbt = microgbtpy.GBT(params)
gbt.train(X_train, y_train, X_valid, y_valid, num_iters,
early_stopping_rounds)
# Predict
for x in X_valid:
pred = gbt.predict(x, gbt.best_iteration())
assert 0 <= pred <= 1
def test_microgbt_boston_rmse():
num_iters = 100
early_stopping_rounds = 10
X_train, X_valid, y_train, y_valid = _load_boston()
# Train
gbt = microgbtpy.GBT(params)
gbt.train(X_train, y_train, X_valid, y_valid, num_iters,
early_stopping_rounds)
# Predict
y_valid_preds = []
for x in X_valid:
y_valid_preds.append(gbt.predict(x, gbt.best_iteration()))
assert mean_squared_error(y_valid, y_valid_preds, squared=False) < 10.0
def test_microgbt_titanic_roc():
num_iters = 100
early_stopping_rounds = 10
X_train, X_valid, y_train, y_valid = load_titanic()
# Train
gbt = microgbtpy.GBT(params)
gbt.train(X_train, y_train, X_valid, y_valid, num_iters,
early_stopping_rounds)
# Predict
y_valid_preds = []
for x in X_valid:
y_valid_preds.append(gbt.predict(x, gbt.best_iteration()))
roc = roc_auc_score(y_valid, y_valid_preds)
assert roc > 0.7, "Area under the curve must be greater than 0.7"
def gbt():
return microgbtpy.GBT(params)
print("Input test valid dimensions {}".format(X_valid.shape))
print("Target valid dims: {}".format(y_valid.shape))
print("Input test dataset dimensions {}".format(X_test.shape))
print("Target test dims: {}".format(y_test.shape))
# Copied from https://github.com/microsoft/LightGBM/blob/master/examples/regression/train.conf
params = {
"gamma": 0.1,
"lambda": 1.0,
"max_depth": 4.0,
"shrinkage_rate": 1.0,
"min_split_gain": 0.1,
"learning_rate": 0.05,
"min_tree_size": 100,
"metric": 1.0
}
# Define the GBT
gbt = microgbtpy.GBT(params)
print(gbt)
# Training related parameters
num_iters = 5
early_stopping_rounds = 5
gbt.train(X_train, y_train, X_valid, y_valid, num_iters, early_stopping_rounds)
logger.info("Best iteration {}".format(gbt.best_iteration()))
regression_metrics(gbt, X_train, y_train, "Training")
regression_metrics(gbt, X_valid, y_valid, "Validation")
regression_metrics(gbt, X_test, y_test, "Testing")