def main():
SpamTrain = "../DataSets/spam.train.txt"
SpamTest = "../DataSets/spam.test.txt"
logging.basicConfig(stream=sys.stderr)
log = logging.getLogger("GradientTreeBoosting.test_on_spam_dataset")
log.setLevel(logging.DEBUG)
data = np.loadtxt(SpamTrain)
train_x, train_y = data[::, 1::], data[::, 0]
ensemble = GradientTreeBoosting(count_steps=200, max_tree_depth=3, step=1e-2, debug=True)
ensemble.fit(train_x, train_y)
ensemble_with_sktree = GradientTreeBoostingViaSklearnTree(count_steps=200, max_tree_depth=3, step=1e-2, debug=True)
ensemble_with_sktree.fit(train_x, train_y)
# for tree in ensemble.models:
# tree.visualize()
# print >>sys.stderr, "\n"
#
# for tree in ensemble_with_sktree.models:
# sklearn_trees.export_graphviz(tree, out_file=sys.stderr)
# print >>sys.stderr, "\n"
sktree = sklearn_trees.DecisionTreeRegressor()
sktree.fit(train_x, train_y)
skensemble = GradientBoostingRegressor(n_estimators=200, max_depth=3)
skensemble.fit(train_x, train_y)
data = np.loadtxt(SpamTest)
test_x, test_y = data[::, 1::], data[::, 0]
prediction = ensemble.predict(test_x)
skprediction = sktree.predict(test_x)
skboosting_prediction = skensemble.predict(test_x)
boosting_with_sktree_prediction = ensemble_with_sktree.predict(test_x)
log.debug("Target: %s" % test_y)
log.debug("Prediction my boosting: %s" % prediction)
log.debug("Prediction boosting with sklearn tree: %s" % boosting_with_sktree_prediction)
log.debug("Mean squared error my boosting: %f" % mean_squared_error(test_y, prediction))
log.debug("Mean squared error sklearn tree: %f" % mean_squared_error(test_y, skprediction))
log.debug("Mean squared error boosting with sklearn tree: %f" %
mean_squared_error(test_y, boosting_with_sktree_prediction))
log.debug("Mean squared error sklearn boosting: %f" % mean_squared_error(test_y, skboosting_prediction))
def test_on_housing_dataset(self):
"""
Test on housing data set
Logging result and MSE for some different models
:return: None
"""
log = logging.getLogger("GradientTreeBoosting.test_on_housing_dataset")
data = np.loadtxt(DecisionTreeTest.HousingDataPath)
x, y = data[::, :-1:], data[::, -1]
kf = KFold(x.shape[0], n_folds=5)
for train, test in kf:
train_x, train_y = x[train], y[train]
test_x, test_y = x[test], y[test]
ensemble = GradientTreeBoosting(count_steps=200, step=1e-2, max_tree_depth=8)
ensemble.fit(train_x, train_y)
ensemble_with_sktree = GradientTreeBoostingViaSklearnTree(count_steps=200, max_tree_depth=8, step=1e-2)
ensemble_with_sktree.fit(train_x, train_y)
sktree = sklearn_trees.DecisionTreeRegressor()
sktree.fit(train_x, train_y)
skensemble = GradientBoostingRegressor()
skensemble.fit(train_x, train_y)
prediction = ensemble.predict(test_x)
skprediction = sktree.predict(test_x)
skboosting_prediction = skensemble.predict(test_x)
boosting_with_sktree_prediction = ensemble_with_sktree.predict(test_x)
log.debug("Target: %s" % test_y)
log.debug("Prediction: %s" % prediction)
log.debug("Mean squared error my boosting: %f" % mean_squared_error(test_y, prediction))
log.debug("Mean squared error sklearn tree: %f" % mean_squared_error(test_y, skprediction))
log.debug("Mean squared error boosting with sklearn tree: %f" %
mean_squared_error(test_y, boosting_with_sktree_prediction))
log.debug("Mean squared error sklearn boosting: %f" % mean_squared_error(test_y, skboosting_prediction))