def main():
# input data file
infile = "./data/batch/df_classification_train.csv"
testfile = "./data/batch/df_classification_test.csv"
# Configure a training object (5 classes)
train_algo = d4p.decision_forest_classification_training(
5,
nTrees=10,
minObservationsInLeafNode=8,
featuresPerNode=3,
engine=d4p.engines_mt19937(seed=777),
varImportance='MDI',
bootstrap=True,
resultsToCompute='computeOutOfBagError')
# Read data. Let's use 3 features per observation
data = read_csv(infile, range(3), t=np.float32)
labels = read_csv(infile, range(3, 4), t=np.float32)
train_result = train_algo.compute(data, labels)
# Traiing result provides (depending on parameters) model, outOfBagError, outOfBagErrorPerObservation and/or variableImportance
# Now let's do some prediction
predict_algo = d4p.decision_forest_classification_prediction(5)
# read test data (with same #features)
pdata = read_csv(testfile, range(3), t=np.float32)
plabels = read_csv(testfile, range(3, 4), t=np.float32)
# now predict using the model from the training above
predict_result = predict_algo.compute(pdata, train_result.model)
# Prediction result provides prediction
assert (predict_result.prediction.shape == (pdata.shape[0], 1))
return (train_result, predict_result, plabels)
def compute(train_data, train_labels, predict_data, method='defaultDense'):
# Configure a training object (5 classes)
train_algo = d4p.decision_forest_classification_training(
5,
fptype='float',
nTrees=10,
minObservationsInLeafNode=8,
featuresPerNode=3,
engine=d4p.engines_mt19937(seed=777),
varImportance='MDI',
bootstrap=True,
resultsToCompute='computeOutOfBagError',
method=method)
# Training result provides (depending on parameters) model,
# outOfBagError, outOfBagErrorPerObservation and/or variableImportance
train_result = train_algo.compute(train_data, train_labels)
# now predict using the model from the training above
predict_algo = d4p.decision_forest_classification_prediction(
nClasses=5,
fptype='float',
resultsToEvaluate="computeClassLabels|computeClassProbabilities",
votingMethod="unweighted")
predict_result = predict_algo.compute(predict_data, train_result.model)
return train_result, predict_result
def _daal_predict(self, X):
X = self._validate_X_predict(X)
if daal_version < (2020,1):
dfc_algorithm = daal4py.decision_forest_classification_prediction(
nClasses = int(self.n_classes_),
fptype = 'float'
)
else:
dfc_algorithm = daal4py.decision_forest_classification_prediction(
nClasses = int(self.n_classes_),
fptype = 'float',
resultsToEvaluate="computeClassLabels"
)
dfc_predictionResult = dfc_algorithm.compute(X, self.daal_model_)
pred = dfc_predictionResult.prediction
return np.take(self.classes_, pred.ravel().astype(np.int, casting='unsafe'))
def df_clsf_predict(X, training_result, n_classes, verbose=False):
algorithm = decision_forest_classification_prediction(
nClasses=n_classes,
fptype='float', # we give float here specifically to match sklearn
)
result = algorithm.compute(X, training_result.model)
return result.prediction
def daal_predict(self, X):
X = self._validate_X_predict(X)
dfc_algorithm = daal4py.decision_forest_classification_prediction(
nClasses=int(self.n_classes_), fptype='float')
dfc_predictionResult = dfc_algorithm.compute(X, self.daal_model_)
pred = dfc_predictionResult.prediction
return np.take(self.classes_,
pred.ravel().astype(np.int, casting='unsafe'))
def _daal_predict_proba(self, X):
X_fptype = getFPType(X)
dfc_algorithm = daal4py.decision_forest_classification_prediction(
nClasses=int(self.n_classes_),
fptype=X_fptype,
resultsToEvaluate="computeClassProbabilities")
dfc_predictionResult = dfc_algorithm.compute(X, self.daal_model_)
pred = dfc_predictionResult.probabilities
return pred
def _daal_predict_proba(self, X):
if not daal_check_version((2021, 'P', 200)):
X = self._validate_X_predict(X)
X_fptype = getFPType(X)
dfc_algorithm = daal4py.decision_forest_classification_prediction(
nClasses=int(self.n_classes_),
fptype=X_fptype,
resultsToEvaluate="computeClassProbabilities")
dfc_predictionResult = dfc_algorithm.compute(X, self.daal_model_)
pred = dfc_predictionResult.probabilities
return pred
def _daal_predict_classifier(self, X):
X_fptype = getFPType(X)
dfc_algorithm = daal4py.decision_forest_classification_prediction(
nClasses=int(self.n_classes_),
fptype=X_fptype,
resultsToEvaluate="computeClassLabels")
if X.shape[1] != self.n_features_in_:
raise ValueError((f'X has {X.shape[1]} features, '
f'but RandomForestClassifier is expecting '
f'{self.n_features_in_} features as input'))
dfc_predictionResult = dfc_algorithm.compute(X, self.daal_model_)
pred = dfc_predictionResult.prediction
return np.take(self.classes_,
pred.ravel().astype(np.int64, casting='unsafe'))
def predict(cls, input):
"""For the input, do the predictions and return them.
Args:
input (a pandas dataframe): The data on which to do the predictions. There will be
one prediction per row in the dataframe"""
with open(param_path, "r") as pf:
params = json.load(pf)
predict_algo = decision_forest_classification_prediction(
int(params["nClasses"]),
fptype=params["fptype"],
method=params["method"],
distributed=(True
if params["distributed"] == "True" else False))
dtype = (np.float64
if params["fptype"] == "double" else np.float32)
clf = cls.get_model()
return predict_algo.compute(input, clf)
