"""
This module implements the base model class. All model things inherit from this class.
"""
import h2o
from . import H2OFrame
from . import H2OVec
from . import H2OTwoDimTable
from . import H2OConnection
[docs]class ModelBase(object):
def __init__(self, dest_key, model_json, metrics_class):
self._key = dest_key
# setup training metrics
if "training_metrics" in model_json["output"]:
tm = model_json["output"]["training_metrics"]
tm = metrics_class(tm,True,False,model_json["algo"])
model_json["output"]["training_metrics"] = tm
# setup validation metrics
if "validation_metrics" in model_json["output"]:
vm = model_json["output"]["validation_metrics"]
if vm is None:
model_json["output"]["validation_metrics"] = None
else:
vm = metrics_class(vm,False,True,model_json["algo"])
model_json["output"]["validation_metrics"] = vm
else:
model_json["output"]["validation_metrics"] = None
self._model_json = model_json
self._metrics_class = metrics_class
def __repr__(self):
self.show()
return ""
[docs] def predict(self, test_data):
"""
Predict on a dataset.
:param test_data: Data to be predicted on.
:return: A new H2OFrame filled with predictions.
"""
if not test_data: raise ValueError("Must specify test data")
# cbind the test_data vecs together and produce a temp key
test_data_key = H2OFrame.send_frame(test_data)
# get the predictions
# this job call is blocking
j = H2OConnection.post_json("Predictions/models/" + self._key + "/frames/" + test_data_key)
# toast the cbound frame
h2o.removeFrameShallow(test_data_key)
# retrieve the prediction frame
prediction_frame_key = j["model_metrics"][0]["predictions"]["frame_id"]["name"]
# get the actual frame meta dta
pred_frame_meta = h2o.frame(prediction_frame_key)["frames"][0]
# toast the prediction frame
h2o.removeFrameShallow(prediction_frame_key)
# collect the vec_ids
vec_ids = pred_frame_meta["vec_ids"]
# get the number of rows
rows = pred_frame_meta["rows"]
# get the column names
cols = [col["label"] for col in pred_frame_meta["columns"]]
# create a set of H2OVec objects
vecs = H2OVec.new_vecs(zip(cols, vec_ids), rows)
# return a new H2OFrame object
return H2OFrame(vecs=vecs)
[docs] def deepfeatures(self, test_data, layer):
"""
Return hidden layer details
:param test_data: Data to create a feature space on
:param layer: 0 index hidden layer
"""
if not test_data: raise ValueError("Must specify test data")
# create test_data by cbinding vecs
test_data_key = H2OFrame.send_frame(test_data)
# get the deepfeatures of the dataset
j = H2OConnection.post_json("Predictions/models/" + self._key + "/frames/" + test_data_key, deep_features_hidden_layer=layer)
# retreive the frame data
deepfeatures_frame_key = j["predictions_frame"]["name"]
df_frame_meta = h2o.frame(deepfeatures_frame_key)["frames"][0]
# create vecs by extracting vec_ids, col length, and col names
vec_ids = df_frame_meta["vec_ids"]
rows = df_frame_meta["rows"]
cols = [col["label"] for col in df_frame_meta["columns"]]
vecs = H2OVec.new_vecs(zip(cols, vec_ids), rows)
# remove test data from kv
h2o.removeFrameShallow(test_data_key)
# finally return frame
return H2OFrame(vecs=vecs)
[docs] def weights(self, matrix_id=0):
"""
Return the frame for the respective weight matrix
:param: matrix_id: an integer, ranging from 0 to number of layers, that specifies the weight matrix to return.
:return: an H2OFrame which represents the weight matrix identified by matrix_id
"""
num_weight_matrices = len(self._model_json['output']['weights'])
if matrix_id not in range(num_weight_matrices):
raise ValueError("Weight matrix does not exist. Model has {0} weight matrices (0-based indexing), but matrix {1} "
"was requested.".format(num_weight_matrices, matrix_id))
j = h2o.frame(self._model_json['output']['weights'][matrix_id]['URL'].split('/')[3])
fr = j['frames'][0]
rows = fr['rows']
vec_ids = fr['vec_ids']
cols = fr['columns']
colnames = [col['label'] for col in cols]
result = H2OFrame(vecs=H2OVec.new_vecs(zip(colnames, vec_ids), rows))
return result
[docs] def biases(self, vector_id=0):
"""
Return the frame for the respective bias vector
:param: vector_id: an integer, ranging from 0 to number of layers, that specifies the bias vector to return.
:return: an H2OFrame which represents the bias vector identified by vector_id
"""
num_bias_vectors = len(self._model_json['output']['biases'])
if vector_id not in range(num_bias_vectors):
raise ValueError("Bias vector does not exist. Model has {0} bias vectors (0-based indexing), but vector {1} "
"was requested.".format(num_bias_vectors, vector_id))
j = h2o.frame(self._model_json['output']['biases'][vector_id]['URL'].split('/')[3])
fr = j['frames'][0]
rows = fr['rows']
vec_ids = fr['vec_ids']
cols = fr['columns']
colnames = [col['label'] for col in cols]
result = H2OFrame(vecs=H2OVec.new_vecs(zip(colnames, vec_ids), rows))
return result
[docs] 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: return model["scoring_history"]
else: print "No score history for this model"
[docs] def summary(self):
"""
Print a detailed summary of the model.
:return:
"""
model = self._model_json["output"]
if model["model_summary"]:
model["model_summary"].show() # H2OTwoDimTable object
[docs] def show(self):
"""
Print innards of model, without regards to type
:return: None
"""
model = self._model_json["output"]
print "Model Details"
print "============="
print self.__class__.__name__, ": ", self._model_json["algo_full_name"]
print "Model Key: ", self._key
self.summary()
print
# training metrics
tm = model["training_metrics"]
if tm: tm.show()
vm = model["validation_metrics"]
if vm: vm.show()
if "scoring_history" in model.keys() and model["scoring_history"]: model["scoring_history"].show()
if "variable_importances" in model.keys() and model["variable_importances"]: model["variable_importances"].show()
[docs] def varimp(self, return_list=False):
"""
Pretty print the variable importances, or return them in a list
:param return_list: if True, then return the variable importances in an list (ordered from most important to least
important). Each entry in the list is a 4-tuple of (variable, relative_importance, scaled_importance, percentage).
:return: None or ordered list
"""
model = self._model_json["output"]
if "variable_importances" in model.keys() and model["variable_importances"]:
if not return_list: return model["variable_importances"].show()
else: return model["variable_importances"].cell_values
else:
print "Warning: This model doesn't have variable importances"
[docs] def residual_deviance(self,train=False,valid=False):
"""
Retreive the residual deviance if this model has the attribute, or None otherwise.
:param train: Get the residual deviance for the training set. If both train and valid are False, then train is selected by default.
:param valid: Get the residual deviance for the validation set. If both train and valid are True, then train is selected by default.
:return: Return the residual deviance, or None if it is not present.
"""
if not train and not valid:
train = True
if train and valid:
train = True
if train:
return self._model_json["output"]["training_metrics"].residual_deviance()
else:
return self._model_json["output"]["validation_metrics"].residual_deviance()
[docs] def residual_degrees_of_freedom(self,train=False,valid=False):
"""
Retreive the residual degress of freedom if this model has the attribute, or None otherwise.
:param train: Get the residual dof for the training set. If both train and valid are False, then train is selected by default.
:param valid: Get the residual dof for the validation set. If both train and valid are True, then train is selected by default.
:return: Return the residual dof, or None if it is not present.
"""
if not train and not valid:
train = True
if train and valid:
train = True
if train:
return self._model_json["output"]["training_metrics"].residual_degrees_of_freedom()
else:
return self._model_json["output"]["validation_metrics"].residual_degrees_of_freedom()
[docs] def null_deviance(self,train=False,valid=False):
"""
Retreive the null deviance if this model has the attribute, or None otherwise.
:param: train Get the null deviance for the training set. If both train and valid are False, then train is selected by default.
:param: valid Get the null deviance for the validation set. If both train and valid are True, then train is selected by default.
:return: Return the null deviance, or None if it is not present.
"""
if not train and not valid:
train = True
if train and valid:
train = True
if train:
return self._model_json["output"]["training_metrics"].null_deviance()
else:
return self._model_json["output"]["validation_metrics"].null_deviance()
[docs] def null_degrees_of_freedom(self,train=False,valid=False):
"""
Retreive the null degress of freedom if this model has the attribute, or None otherwise.
:param train: Get the null dof for the training set. If both train and valid are False, then train is selected by default.
:param valid: Get the null dof for the validation set. If both train and valid are True, then train is selected by default.
:return: Return the null dof, or None if it is not present.
"""
if not train and not valid:
train = True
if train and valid:
train = True
if train:
return self._model_json["output"]["training_metrics"].null_degrees_of_freedom()
else:
return self._model_json["output"]["validation_metrics"].null_degrees_of_freedom()
[docs] def pprint_coef(self):
"""
Pretty print the coefficents table (includes normalized coefficients)
:return: None
"""
print self._model_json["output"]["coefficients_table"] # will return None if no coefs!
[docs] def coef(self):
"""
:return: Return the coefficients for this model.
"""
tbl = self._model_json["output"]["coefficients_table"]
if tbl is None: return None
tbl = tbl.cell_values
return {a[0]:a[1] for a in tbl}
[docs] def coef_norm(self):
"""
:return: Return the normalized coefficients
"""
tbl = self._model_json["output"]["coefficients_table"]
if tbl is None: return None
tbl = tbl.cell_values
return {a[0]:a[2] for a in tbl}
[docs] def r2(self, train=False, valid=False):
"""
Return the R^2 for this regression model.
The R^2 value is defined to be 1 - MSE/var,
where var is computed as sigma*sigma.
:param train: If train is True, then return the R^2 value for the training data. If train and valid are both False, then return the training R^2.
:param valid: If valid is True, then return the R^2 value for the validation data. If train and valid are both True, then return the validation R^2.
:return: The R^2 for this regression model.
"""
tm = ModelBase._get_metrics(self, *ModelBase._train_or_valid(train,valid))
if tm is None: return None
return tm.r2()
[docs] def mse(self, train=False,valid=False):
"""
:param train: If train is True, then return the MSE value for the training data. If train and valid are both False, then return the training MSE.
:param valid: If valid is True, then return the MSE value for the validation data. If train and valid are both True, then return the validation MSE.
:return: The MSE for this regression model.
"""
tm = ModelBase._get_metrics(self, *ModelBase._train_or_valid(train,valid))
if tm is None: return None
return tm.mse()
[docs] def logloss(self, train=False, valid=False):
"""
Get the Log Loss.
If both train and valid are False, return the train.
If both train and valid are True, return the valid.
:param train: Return the log loss for training data.
:param valid: Return the log loss for the validation data.
:return: Retrieve the log loss coefficient for this set of metrics
"""
tm = ModelBase._get_metrics(self,*ModelBase._train_or_valid(train, valid))
if tm is None: return None
return tm.logloss()
[docs] def auc(self, train=False, valid=False):
"""
Get the AUC.
If both train and valid are False, return the train.
If both train and valid are True, return the valid.
:param train: Return the AUC for training data.
:param valid: Return the AUC for the validation data.
:return: Retrieve the AUC coefficient for this set of metrics
"""
tm = ModelBase._get_metrics(self,*ModelBase._train_or_valid(train, valid))
if tm is None: return None
tm = tm._metric_json
return tm.auc()
[docs] def aic(self, train=False, valid=False):
"""
Get the AIC.
If both train and valid are False, return the train.
If both train and valid are True, return the valid.
:param train: Return the AIC for training data.
:param valid: Return the AIC for the validation data.
:return: Retrieve the AIC for this set of metrics
"""
tm = ModelBase._get_metrics(self,*ModelBase._train_or_valid(train, valid))
if tm is None: return None
tm = tm._metric_json
return tm["AIC"]
[docs] def giniCoef(self, train=False, valid=False):
"""
Get the Gini.
If both train and valid are False, return the train.
If both train and valid are True, return the valid.
:param train: Return the Gini for training data.
:param valid: Return the Gini for the validation data.
:return: Retrieve the Gini coefficient for this set of metrics
"""
tm = ModelBase._get_metrics(self, *ModelBase._train_or_valid(train, valid))
if tm is None: return None
tm = tm._metric_json
return tm.giniCoef()
[docs] def download_pojo(self,path=""):
"""
Download the POJO for this model to the directory specified by path (no trailing slash!).
If path is "", then dump to screen.
:param model: Retrieve this model's scoring POJO.
:param path: An absolute path to the directory where POJO should be saved.
:return: None
"""
h2o.download_pojo(self,path) # call the "package" function
@staticmethod
def _get_metrics(o, train, valid):
if train:
return o._model_json["output"]["training_metrics"]
if valid:
return o._model_json["output"]["validation_metrics"]
raise ValueError("`_get_metrics` demands `train` or `valid` to be True.")
@staticmethod
def _train_or_valid(train,valid):
"""
Internal static method.
:param train: a boolean for train. Ignored, however.
:param valid: a boolean for valid
:return: true if train, false if valid. If both are false, return True for train.
"""
if valid: return [False, True]
return [True,False]
# Delete from cluster as model goes out of scope
def __del__(self):
h2o.remove(self._key)
@staticmethod
def _has(dictionary, key):
return key in dictionary and dictionary[key] is not None