def varimp(self, use_pandas=False):
"""
Pretty print the variable importances, or return them in a list
Parameters
----------
use_pandas: boolean, optional
If True, then the variable importances will be returned as a pandas data frame.
Returns
-------
A list or Pandas DataFrame.
"""
model = self._model_json["output"]
if "variable_importances" in model.keys(
) and model["variable_importances"]:
vals = model["variable_importances"].cell_values
header = model["variable_importances"].col_header
if use_pandas and h2o.can_use_pandas():
import pandas
return pandas.DataFrame(vals, columns=header)
else:
return vals
else:
print "Warning: This model doesn't have variable importances"
def varimp(self, use_pandas=False):
"""
Pretty print the variable importances, or return them in a list
Parameters
----------
use_pandas: boolean, optional
If True, then the variable importances will be returned as a pandas data frame.
Returns
-------
A list or Pandas DataFrame.
"""
model = self._model_json["output"]
if "variable_importances" in model.keys(
) and model["variable_importances"]:
vals = model["variable_importances"].cell_values
header = model["variable_importances"].col_header
if use_pandas and h2o.can_use_pandas():
import pandas
return pandas.DataFrame(vals, columns=header)
else:
return vals
else:
print "Warning: This model doesn't have variable importances"
def show(self, header=True):
if h2o.can_use_pandas():
import pandas
pandas.options.display.max_rows = 20
print pandas.DataFrame(self.cell_values, columns=self.col_header)
return
print
if header: print self.table_header + ":"
print
table = copy.deepcopy(self.cell_values)
nr = 0
if _is_list_of_lists(table):
nr = len(
table
) # only set if we truly have multiple rows... not just one long row :)
if nr > 20: # create a truncated view of the table, first/last 5 rows
trunc_table = []
trunc_table += [v for v in table[:5]]
trunc_table.append(["---"] * len(table[0]))
trunc_table += [v for v in table[(nr - 5):]]
table = trunc_table
h2o.H2ODisplay(table,
self.col_header,
numalign="left",
stralign="left")
def score_history(self):
"""
Retrieve Model Score History
:return: the score history (H2OTwoDimTable)
"""
model = self._model_json["output"]
if 'scoring_history' in model.keys() and model["scoring_history"] != None:
s = model["scoring_history"]
if h2o.can_use_pandas():
import pandas
pandas.options.display.max_rows = 20
return pandas.DataFrame(s.cell_values,columns=s.col_header)
return s
else: print "No score history for this model"
def score_history(self):
"""
Retrieve Model Score History
:return: the score history (H2OTwoDimTable)
"""
model = self._model_json["output"]
if 'scoring_history' in model.keys() and model["scoring_history"] != None:
s = model["scoring_history"]
if h2o.can_use_pandas():
import pandas
pandas.options.display.max_rows = 20
return pandas.DataFrame(s.cell_values,columns=s.col_header)
return model["scoring_history"]
else: print "No score history for this model"
def show(self, header=True):
if h2o.can_use_pandas():
import pandas
pandas.options.display.max_rows = 20
print pandas.DataFrame(self.cell_values,columns=self.col_header)
return
print
if header:
print self.table_header + ":",
if self.table_description: print self.table_description
print
table = copy.deepcopy(self.cell_values)
nr=0
if _is_list_of_lists(table): nr = len(table) # only set if we truly have multiple rows... not just one long row :)
if nr > 20: # create a truncated view of the table, first/last 5 rows
trunc_table =[]
trunc_table += [ v for v in table[:5]]
trunc_table.append(["---"]*len(table[0]))
trunc_table += [v for v in table[(nr-5):]]
table = trunc_table
h2o.H2ODisplay(table, self.col_header, numalign="left", stralign="left")
def as_data_frame(self, use_pandas=True):
"""Obtain the dataset as a python-local object.
:param use_pandas: A flag specifying whether or not to return a pandas DataFrame.
:return: A local python object (a list of lists of strings, each list is a row, if use_pandas=False, otherwise a
pandas DataFrame) containing this self._newExpr instance's data.
"""
url = 'http://' + h2o.H2OConnection.ip() + ':' + str(h2o.H2OConnection.port()) + "/3/DownloadDataset?frame_id=" + urllib.quote(self.frame_id) + "&hex_string=false"
response = urllib2.urlopen(url)
if h2o.can_use_pandas() and use_pandas:
import pandas
df = pandas.read_csv(response,low_memory=False)
time_cols = []
category_cols = []
if self.types is not None:
for col_name in self.names:
xtype = self.type(col_name)
if xtype.lower() == 'time': time_cols.append(col_name)
elif xtype.lower() == 'enum': category_cols.append(col_name)
#change Time to pandas datetime
if time_cols:
#hacky way to get the utc offset
from datetime import datetime
sample_timestamp = 1380610868
utc_offset = 1000 * ((datetime.utcfromtimestamp(sample_timestamp) - datetime.fromtimestamp(sample_timestamp)).total_seconds())
try:
df[time_cols] = (df[time_cols] - utc_offset).astype('datetime64[ms]')
except pandas.tslib.OutOfBoundsDatetime:
pass
#change Enum to pandas category
for cat_col in category_cols: #for loop is required
df[cat_col] = df[cat_col].astype('category')
return df
else:
cr = csv.reader(response)
t_col_list = [[''] if row == [] else row for row in cr]
return [list(x) for x in zip(*t_col_list)]
def _plot(self, timestep, metric, **kwargs):
# check for matplotlib. exit if absent
try:
imp.find_module('matplotlib')
import matplotlib
if 'server' in kwargs.keys() and kwargs['server']: matplotlib.use('Agg', warn=False)
import matplotlib.pyplot as plt
except ImportError:
print "matplotlib is required for this function!"
return
scoring_history = self.score_history()
# Separate functionality for GLM since its output is different from other algos
if self._model_json["algo"] == "glm":
# GLM has only one timestep option, which is `iteration`
timestep = "iteration"
if metric == "AUTO": metric = "log_likelihood"
elif metric not in ("log_likelihood", "objective"):
raise ValueError("for GLM, metric must be one of: log_likelihood, objective")
plt.xlabel(timestep)
plt.ylabel(metric)
plt.title("Validation Scoring History")
plt.plot(scoring_history[timestep], scoring_history[metric])
elif self._model_json["algo"] in ("deeplearning", "drf", "gbm"):
# Set timestep
if self._model_json["algo"] in ("gbm", "drf"):
if timestep == "AUTO": timestep = "number_of_trees"
elif timestep not in ("duration","number_of_trees"):
raise ValueError("timestep for gbm or drf must be one of: duration, number_of_trees")
else: #self._model_json["algo"] == "deeplearning":
# Delete first row of DL scoring history since it contains NAs & NaNs
if scoring_history["samples"][0] == 0:
scoring_history = scoring_history[1:]
if timestep == "AUTO": timestep = "epochs"
elif timestep not in ("epochs","samples","duration"):
raise ValueError("timestep for deeplearning must be one of: epochs, samples, duration")
training_metric = "training_{}".format(metric)
validation_metric = "validation_{}".format(metric)
if timestep == "duration":
dur_colname = "duration_{}".format(scoring_history["duration"][1].split()[1])
scoring_history[dur_colname] = map(lambda x: str(x).split()[0],scoring_history["duration"])
timestep = dur_colname
if h2o.can_use_pandas():
valid = validation_metric in list(scoring_history)
ylim = (scoring_history[[training_metric, validation_metric]].min().min(), scoring_history[[training_metric, validation_metric]].max().max()) if valid \
else (scoring_history[training_metric].min(), scoring_history[training_metric].max())
else:
valid = validation_metric in scoring_history.col_header
ylim = (min(min(scoring_history[[training_metric, validation_metric]])), max(max(scoring_history[[training_metric, validation_metric]]))) if valid \
else (min(scoring_history[training_metric]), max(scoring_history[training_metric]))
if valid: #Training and validation scoring history
plt.xlabel(timestep)
plt.ylabel(metric)
plt.title("Scoring History")
plt.ylim(ylim)
plt.plot(scoring_history[timestep], scoring_history[training_metric], label = "Training")
plt.plot(scoring_history[timestep], scoring_history[validation_metric], color = "orange", label = "Validation")
plt.legend()
else: #Training scoring history only
plt.xlabel(timestep)
plt.ylabel(training_metric)
plt.title("Training Scoring History")
plt.ylim(ylim)
plt.plot(scoring_history[timestep], scoring_history[training_metric])
else: # algo is not glm, deeplearning, drf, gbm
raise ValueError("Plotting not implemented for this type of model")
if "server" not in kwargs.keys() or not kwargs["server"]: plt.show()
def _plot(self, timestep, metric, **kwargs):
# check for matplotlib. exit if absent
try:
imp.find_module('matplotlib')
import matplotlib
if 'server' in kwargs.keys() and kwargs['server']: matplotlib.use('Agg', warn=False)
import matplotlib.pyplot as plt
except ImportError:
print "matplotlib is required for this function!"
return
scoring_history = self.score_history()
# Separate functionality for GLM since its output is different from other algos
if self._model_json["algo"] == "glm":
# GLM has only one timestep option, which is `iteration`
timestep = "iteration"
if metric == "AUTO": metric = "log_likelihood"
elif metric not in ("log_likelihood", "objective"):
raise ValueError("for GLM, metric must be one of: log_likelihood, objective")
plt.xlabel(timestep)
plt.ylabel(metric)
plt.title("Validation Scoring History")
plt.plot(scoring_history[timestep], scoring_history[metric])
elif self._model_json["algo"] in ("deeplearning", "drf", "gbm"):
# Set timestep
if self._model_json["algo"] in ("gbm", "drf"):
if timestep == "AUTO": timestep = "number_of_trees"
elif timestep not in ("duration","number_of_trees"):
raise ValueError("timestep for gbm or drf must be one of: duration, number_of_trees")
else: #self._model_json["algo"] == "deeplearning":
# Delete first row of DL scoring history since it contains NAs & NaNs
if scoring_history["samples"][0] == 0:
scoring_history = scoring_history[1:]
if timestep == "AUTO": timestep = "epochs"
elif timestep not in ("epochs","samples","duration"):
raise ValueError("timestep for deeplearning must be one of: epochs, samples, duration")
training_metric = "training_{}".format(metric)
validation_metric = "validation_{}".format(metric)
if timestep == "duration":
dur_colname = "duration_{}".format(scoring_history["duration"][1].split()[1])
scoring_history[dur_colname] = map(lambda x: str(x).split()[0],scoring_history["duration"])
timestep = dur_colname
if h2o.can_use_pandas():
valid = validation_metric in list(scoring_history)
ylim = (scoring_history[[training_metric, validation_metric]].min().min(), scoring_history[[training_metric, validation_metric]].max().max()) if valid \
else (scoring_history[training_metric].min(), scoring_history[training_metric].max())
else:
valid = validation_metric in scoring_history.col_header
ylim = (min(min(scoring_history[[training_metric, validation_metric]])), max(max(scoring_history[[training_metric, validation_metric]]))) if valid \
else (min(scoring_history[training_metric]), max(scoring_history[training_metric]))
if valid: #Training and validation scoring history
plt.xlabel(timestep)
plt.ylabel(metric)
plt.title("Scoring History")
plt.ylim(ylim)
plt.plot(scoring_history[timestep], scoring_history[training_metric], label = "Training")
plt.plot(scoring_history[timestep], scoring_history[validation_metric], color = "orange", label = "Validation")
plt.legend()
else: #Training scoring history only
plt.xlabel(timestep)
plt.ylabel(training_metric)
plt.title("Training Scoring History")
plt.ylim(ylim)
plt.plot(scoring_history[timestep], scoring_history[training_metric])
else: # algo is not glm, deeplearning, drf, gbm
raise ValueError("Plotting not implemented for this type of model")
if "server" not in kwargs.keys() or not kwargs["server"]: plt.show()