Coverage for pyspark/mllib/classification.py : 92%

# 

# Licensed to the Apache Software Foundation (ASF) under one or more 

# contributor license agreements. See the NOTICE file distributed with 

# this work for additional information regarding copyright ownership. 

# The ASF licenses this file to You under the Apache License, Version 2.0 

# (the "License"); you may not use this file except in compliance with 

# the License. You may obtain a copy of the License at 

# 

# http://www.apache.org/licenses/LICENSE-2.0 

# 

# Unless required by applicable law or agreed to in writing, software 

# distributed under the License is distributed on an "AS IS" BASIS, 

# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 

# See the License for the specific language governing permissions and 

# limitations under the License. 

# 

from math import exp 

import sys 

import warnings 

import numpy 

from pyspark import RDD, since 

from pyspark.mllib.common import callMLlibFunc, _py2java, _java2py 

from pyspark.mllib.linalg import _convert_to_vector 

from pyspark.mllib.regression import ( 

LabeledPoint, LinearModel, _regression_train_wrapper, 

StreamingLinearAlgorithm) 

from pyspark.mllib.util import Saveable, Loader, inherit_doc 

__all__ = ['LogisticRegressionModel', 'LogisticRegressionWithSGD', 'LogisticRegressionWithLBFGS', 

'SVMModel', 'SVMWithSGD', 'NaiveBayesModel', 'NaiveBayes', 

'StreamingLogisticRegressionWithSGD'] 

class LinearClassificationModel(LinearModel): 

""" 

A private abstract class representing a multiclass classification 

model. The categories are represented by int values: 0, 1, 2, etc. 

""" 

def __init__(self, weights, intercept): 

super(LinearClassificationModel, self).__init__(weights, intercept) 

self._threshold = None 

@since('1.4.0') 

def setThreshold(self, value): 

""" 

Sets the threshold that separates positive predictions from 

negative predictions. An example with prediction score greater 

than or equal to this threshold is identified as a positive, 

and negative otherwise. It is used for binary classification 

only. 

""" 

self._threshold = value 

@property 

@since('1.4.0') 

def threshold(self): 

""" 

Returns the threshold (if any) used for converting raw 

prediction scores into 0/1 predictions. It is used for 

binary classification only. 

""" 

return self._threshold 

@since('1.4.0') 

def clearThreshold(self): 

""" 

Clears the threshold so that `predict` will output raw 

prediction scores. It is used for binary classification only. 

""" 

self._threshold = None 

@since('1.4.0') 

def predict(self, test): 

""" 

Predict values for a single data point or an RDD of points 

using the model trained. 

""" 

raise NotImplementedError 

class LogisticRegressionModel(LinearClassificationModel): 

""" 

Classification model trained using Multinomial/Binary Logistic 

Regression. 

.. versionadded:: 0.9.0 

Parameters 

---------- 

weights : :py:class:`pyspark.mllib.linalg.Vector` 

Weights computed for every feature. 

intercept : float 

Intercept computed for this model. (Only used in Binary Logistic 

Regression. In Multinomial Logistic Regression, the intercepts will 

not be a single value, so the intercepts will be part of the 

weights.) 

numFeatures : int 

The dimension of the features. 

numClasses : int 

The number of possible outcomes for k classes classification problem 

in Multinomial Logistic Regression. By default, it is binary 

logistic regression so numClasses will be set to 2. 

Examples 

-------- 

>>> from pyspark.mllib.linalg import SparseVector 

>>> data = [ 

... LabeledPoint(0.0, [0.0, 1.0]), 

... LabeledPoint(1.0, [1.0, 0.0]), 

... ] 

>>> lrm = LogisticRegressionWithSGD.train(sc.parallelize(data), iterations=10) 

>>> lrm.predict([1.0, 0.0]) 

1 

>>> lrm.predict([0.0, 1.0]) 

0 

>>> lrm.predict(sc.parallelize([[1.0, 0.0], [0.0, 1.0]])).collect() 

[1, 0] 

>>> lrm.clearThreshold() 

>>> lrm.predict([0.0, 1.0]) 

0.279... 

>>> sparse_data = [ 

... LabeledPoint(0.0, SparseVector(2, {0: 0.0})), 

... LabeledPoint(1.0, SparseVector(2, {1: 1.0})), 

... LabeledPoint(0.0, SparseVector(2, {0: 1.0})), 

... LabeledPoint(1.0, SparseVector(2, {1: 2.0})) 

... ] 

>>> lrm = LogisticRegressionWithSGD.train(sc.parallelize(sparse_data), iterations=10) 

>>> lrm.predict(numpy.array([0.0, 1.0])) 

1 

>>> lrm.predict(numpy.array([1.0, 0.0])) 

0 

>>> lrm.predict(SparseVector(2, {1: 1.0})) 

1 

>>> lrm.predict(SparseVector(2, {0: 1.0})) 

0 

>>> import os, tempfile 

>>> path = tempfile.mkdtemp() 

>>> lrm.save(sc, path) 

>>> sameModel = LogisticRegressionModel.load(sc, path) 

>>> sameModel.predict(numpy.array([0.0, 1.0])) 

1 

>>> sameModel.predict(SparseVector(2, {0: 1.0})) 

0 

>>> from shutil import rmtree 

>>> try: 

... rmtree(path) 

... except: 

... pass 

>>> multi_class_data = [ 

... LabeledPoint(0.0, [0.0, 1.0, 0.0]), 

... LabeledPoint(1.0, [1.0, 0.0, 0.0]), 

... LabeledPoint(2.0, [0.0, 0.0, 1.0]) 

... ] 

>>> data = sc.parallelize(multi_class_data) 

>>> mcm = LogisticRegressionWithLBFGS.train(data, iterations=10, numClasses=3) 

>>> mcm.predict([0.0, 0.5, 0.0]) 

0 

>>> mcm.predict([0.8, 0.0, 0.0]) 

1 

>>> mcm.predict([0.0, 0.0, 0.3]) 

2 

""" 

def __init__(self, weights, intercept, numFeatures, numClasses): 

super(LogisticRegressionModel, self).__init__(weights, intercept) 

self._numFeatures = int(numFeatures) 

self._numClasses = int(numClasses) 

self._threshold = 0.5 

if self._numClasses == 2: 

self._dataWithBiasSize = None 

self._weightsMatrix = None 

else: 

self._dataWithBiasSize = self._coeff.size // (self._numClasses - 1) 

self._weightsMatrix = self._coeff.toArray().reshape(self._numClasses - 1, 

self._dataWithBiasSize) 

@property 

@since('1.4.0') 

def numFeatures(self): 

""" 

Dimension of the features. 

""" 

return self._numFeatures 

@property 

@since('1.4.0') 

def numClasses(self): 

""" 

Number of possible outcomes for k classes classification problem 

in Multinomial Logistic Regression. 

""" 

return self._numClasses 

@since('0.9.0') 

def predict(self, x): 

""" 

Predict values for a single data point or an RDD of points 

using the model trained. 

""" 

if isinstance(x, RDD): 

return x.map(lambda v: self.predict(v)) 

x = _convert_to_vector(x) 

if self.numClasses == 2: 

margin = self.weights.dot(x) + self._intercept 

if margin > 0: 

prob = 1 / (1 + exp(-margin)) 

else: 

exp_margin = exp(margin) 

prob = exp_margin / (1 + exp_margin) 

if self._threshold is None: 

return prob 

else: 

return 1 if prob > self._threshold else 0 

else: 

best_class = 0 

max_margin = 0.0 

223 ↛ 224line 223 didn't jump to line 224, because the condition on line 223 was never true if x.size + 1 == self._dataWithBiasSize: 

for i in range(0, self._numClasses - 1): 

margin = x.dot(self._weightsMatrix[i][0:x.size]) + \ 

self._weightsMatrix[i][x.size] 

if margin > max_margin: 

max_margin = margin 

best_class = i + 1 

else: 

for i in range(0, self._numClasses - 1): 

margin = x.dot(self._weightsMatrix[i]) 

if margin > max_margin: 

max_margin = margin 

best_class = i + 1 

return best_class 

@since('1.4.0') 

def save(self, sc, path): 

""" 

Save this model to the given path. 

""" 

java_model = sc._jvm.org.apache.spark.mllib.classification.LogisticRegressionModel( 

_py2java(sc, self._coeff), self.intercept, self.numFeatures, self.numClasses) 

java_model.save(sc._jsc.sc(), path) 

@classmethod 

@since('1.4.0') 

def load(cls, sc, path): 

""" 

Load a model from the given path. 

""" 

java_model = sc._jvm.org.apache.spark.mllib.classification.LogisticRegressionModel.load( 

sc._jsc.sc(), path) 

weights = _java2py(sc, java_model.weights()) 

intercept = java_model.intercept() 

numFeatures = java_model.numFeatures() 

numClasses = java_model.numClasses() 

threshold = java_model.getThreshold().get() 

model = LogisticRegressionModel(weights, intercept, numFeatures, numClasses) 

model.setThreshold(threshold) 

return model 

def __repr__(self): 

return self._call_java("toString") 

class LogisticRegressionWithSGD(object): 

""" 

Train a classification model for Binary Logistic Regression using Stochastic Gradient Descent. 

.. versionadded:: 0.9.0 

.. deprecated:: 2.0.0 

Use ml.classification.LogisticRegression or LogisticRegressionWithLBFGS. 

""" 

@classmethod 

def train(cls, data, iterations=100, step=1.0, miniBatchFraction=1.0, 

initialWeights=None, regParam=0.01, regType="l2", intercept=False, 

validateData=True, convergenceTol=0.001): 

""" 

Train a logistic regression model on the given data. 

.. versionadded:: 0.9.0 

Parameters 

---------- 

data : :py:class:`pyspark.RDD` 

The training data, an RDD of :py:class:`pyspark.mllib.regression.LabeledPoint`. 

iterations : int, optional 

The number of iterations. 

(default: 100) 

step : float, optional 

The step parameter used in SGD. 

(default: 1.0) 

miniBatchFraction : float, optional 

Fraction of data to be used for each SGD iteration. 

(default: 1.0) 

initialWeights : :py:class:`pyspark.mllib.linalg.Vector` or convertible, optional 

The initial weights. 

(default: None) 

regParam : float, optional 

The regularizer parameter. 

(default: 0.01) 

regType : str, optional 

The type of regularizer used for training our model. 

Supported values: 

- "l1" for using L1 regularization 

- "l2" for using L2 regularization (default) 

- None for no regularization 

intercept : bool, optional 

Boolean parameter which indicates the use or not of the 

augmented representation for training data (i.e., whether bias 

features are activated or not). 

(default: False) 

validateData : bool, optional 

Boolean parameter which indicates if the algorithm should 

validate data before training. 

(default: True) 

convergenceTol : float, optional 

A condition which decides iteration termination. 

(default: 0.001) 

""" 

warnings.warn( 

"Deprecated in 2.0.0. Use ml.classification.LogisticRegression or " 

"LogisticRegressionWithLBFGS.", FutureWarning) 

def train(rdd, i): 

return callMLlibFunc("trainLogisticRegressionModelWithSGD", rdd, int(iterations), 

float(step), float(miniBatchFraction), i, float(regParam), regType, 

bool(intercept), bool(validateData), float(convergenceTol)) 

return _regression_train_wrapper(train, LogisticRegressionModel, data, initialWeights) 

class LogisticRegressionWithLBFGS(object): 

""" 

Train a classification model for Multinomial/Binary Logistic Regression 

using Limited-memory BFGS. 

Standard feature scaling and L2 regularization are used by default. 

.. versionadded:: 1.2.0 

""" 

@classmethod 

def train(cls, data, iterations=100, initialWeights=None, regParam=0.0, regType="l2", 

intercept=False, corrections=10, tolerance=1e-6, validateData=True, numClasses=2): 

""" 

Train a logistic regression model on the given data. 

.. versionadded:: 1.2.0 

Parameters 

---------- 

data : :py:class:`pyspark.RDD` 

The training data, an RDD of :py:class:`pyspark.mllib.regression.LabeledPoint`. 

iterations : int, optional 

The number of iterations. 

(default: 100) 

initialWeights : :py:class:`pyspark.mllib.linalg.Vector` or convertible, optional 

The initial weights. 

(default: None) 

regParam : float, optional 

The regularizer parameter. 

(default: 0.01) 

regType : str, optional 

The type of regularizer used for training our model. 

Supported values: 

- "l1" for using L1 regularization 

- "l2" for using L2 regularization (default) 

- None for no regularization 

intercept : bool, optional 

Boolean parameter which indicates the use or not of the 

augmented representation for training data (i.e., whether bias 

features are activated or not). 

(default: False) 

corrections : int, optional 

The number of corrections used in the LBFGS update. 

If a known updater is used for binary classification, 

it calls the ml implementation and this parameter will 

have no effect. (default: 10) 

tolerance : float, optional 

The convergence tolerance of iterations for L-BFGS. 

(default: 1e-6) 

validateData : bool, optional 

Boolean parameter which indicates if the algorithm should 

validate data before training. 

(default: True) 

numClasses : int, optional 

The number of classes (i.e., outcomes) a label can take in 

Multinomial Logistic Regression. 

(default: 2) 

Examples 

-------- 

>>> data = [ 

... LabeledPoint(0.0, [0.0, 1.0]), 

... LabeledPoint(1.0, [1.0, 0.0]), 

... ] 

>>> lrm = LogisticRegressionWithLBFGS.train(sc.parallelize(data), iterations=10) 

>>> lrm.predict([1.0, 0.0]) 

1 

>>> lrm.predict([0.0, 1.0]) 

0 

""" 

def train(rdd, i): 

return callMLlibFunc("trainLogisticRegressionModelWithLBFGS", rdd, int(iterations), i, 

float(regParam), regType, bool(intercept), int(corrections), 

float(tolerance), bool(validateData), int(numClasses)) 

413 ↛ 421line 413 didn't jump to line 421, because the condition on line 413 was never false if initialWeights is None: 

if numClasses == 2: 

initialWeights = [0.0] * len(data.first().features) 

else: 

417 ↛ 418line 417 didn't jump to line 418, because the condition on line 417 was never true if intercept: 

initialWeights = [0.0] * (len(data.first().features) + 1) * (numClasses - 1) 

else: 

initialWeights = [0.0] * len(data.first().features) * (numClasses - 1) 

return _regression_train_wrapper(train, LogisticRegressionModel, data, initialWeights) 

class SVMModel(LinearClassificationModel): 

""" 

Model for Support Vector Machines (SVMs). 

.. versionadded:: 0.9.0 

Parameters 

---------- 

weights : :py:class:`pyspark.mllib.linalg.Vector` 

Weights computed for every feature. 

intercept : float 

Intercept computed for this model. 

Examples 

-------- 

>>> from pyspark.mllib.linalg import SparseVector 

>>> data = [ 

... LabeledPoint(0.0, [0.0]), 

... LabeledPoint(1.0, [1.0]), 

... LabeledPoint(1.0, [2.0]), 

... LabeledPoint(1.0, [3.0]) 

... ] 

>>> svm = SVMWithSGD.train(sc.parallelize(data), iterations=10) 

>>> svm.predict([1.0]) 

1 

>>> svm.predict(sc.parallelize([[1.0]])).collect() 

[1] 

>>> svm.clearThreshold() 

>>> svm.predict(numpy.array([1.0])) 

1.44... 

>>> sparse_data = [ 

... LabeledPoint(0.0, SparseVector(2, {0: -1.0})), 

... LabeledPoint(1.0, SparseVector(2, {1: 1.0})), 

... LabeledPoint(0.0, SparseVector(2, {0: 0.0})), 

... LabeledPoint(1.0, SparseVector(2, {1: 2.0})) 

... ] 

>>> svm = SVMWithSGD.train(sc.parallelize(sparse_data), iterations=10) 

>>> svm.predict(SparseVector(2, {1: 1.0})) 

1 

>>> svm.predict(SparseVector(2, {0: -1.0})) 

0 

>>> import os, tempfile 

>>> path = tempfile.mkdtemp() 

>>> svm.save(sc, path) 

>>> sameModel = SVMModel.load(sc, path) 

>>> sameModel.predict(SparseVector(2, {1: 1.0})) 

1 

>>> sameModel.predict(SparseVector(2, {0: -1.0})) 

0 

>>> from shutil import rmtree 

>>> try: 

... rmtree(path) 

... except: 

... pass 

""" 

def __init__(self, weights, intercept): 

super(SVMModel, self).__init__(weights, intercept) 

self._threshold = 0.0 

@since('0.9.0') 

def predict(self, x): 

""" 

Predict values for a single data point or an RDD of points 

using the model trained. 

""" 

if isinstance(x, RDD): 

return x.map(lambda v: self.predict(v)) 

x = _convert_to_vector(x) 

margin = self.weights.dot(x) + self.intercept 

if self._threshold is None: 

return margin 

else: 

return 1 if margin > self._threshold else 0 

@since('1.4.0') 

def save(self, sc, path): 

""" 

Save this model to the given path. 

""" 

java_model = sc._jvm.org.apache.spark.mllib.classification.SVMModel( 

_py2java(sc, self._coeff), self.intercept) 

java_model.save(sc._jsc.sc(), path) 

@classmethod 

@since('1.4.0') 

def load(cls, sc, path): 

""" 

Load a model from the given path. 

""" 

java_model = sc._jvm.org.apache.spark.mllib.classification.SVMModel.load( 

sc._jsc.sc(), path) 

weights = _java2py(sc, java_model.weights()) 

intercept = java_model.intercept() 

threshold = java_model.getThreshold().get() 

model = SVMModel(weights, intercept) 

model.setThreshold(threshold) 

return model 

class SVMWithSGD(object): 

""" 

Train a Support Vector Machine (SVM) using Stochastic Gradient Descent. 

.. versionadded:: 0.9.0 

""" 

@classmethod 

def train(cls, data, iterations=100, step=1.0, regParam=0.01, 

miniBatchFraction=1.0, initialWeights=None, regType="l2", 

intercept=False, validateData=True, convergenceTol=0.001): 

""" 

Train a support vector machine on the given data. 

.. versionadded:: 0.9.0 

Parameters 

---------- 

data : :py:class:`pyspark.RDD` 

The training data, an RDD of :py:class:`pyspark.mllib.regression.LabeledPoint`. 

iterations : int, optional 

The number of iterations. 

(default: 100) 

step : float, optional 

The step parameter used in SGD. 

(default: 1.0) 

regParam : float, optional 

The regularizer parameter. 

(default: 0.01) 

miniBatchFraction : float, optional 

Fraction of data to be used for each SGD iteration. 

(default: 1.0) 

initialWeights : :py:class:`pyspark.mllib.linalg.Vector` or convertible, optional 

The initial weights. 

(default: None) 

regType : str, optional 

The type of regularizer used for training our model. 

Allowed values: 

- "l1" for using L1 regularization 

- "l2" for using L2 regularization (default) 

- None for no regularization 

intercept : bool, optional 

Boolean parameter which indicates the use or not of the 

augmented representation for training data (i.e. whether bias 

features are activated or not). 

(default: False) 

validateData : bool, optional 

Boolean parameter which indicates if the algorithm should 

validate data before training. 

(default: True) 

convergenceTol : float, optional 

A condition which decides iteration termination. 

(default: 0.001) 

""" 

def train(rdd, i): 

return callMLlibFunc("trainSVMModelWithSGD", rdd, int(iterations), float(step), 

float(regParam), float(miniBatchFraction), i, regType, 

bool(intercept), bool(validateData), float(convergenceTol)) 

return _regression_train_wrapper(train, SVMModel, data, initialWeights) 

@inherit_doc 

class NaiveBayesModel(Saveable, Loader): 

""" 

Model for Naive Bayes classifiers. 

.. versionadded:: 0.9.0 

Parameters 

---------- 

labels : :py:class:`numpy.ndarray` 

List of labels. 

pi : :py:class:`numpy.ndarray` 

Log of class priors, whose dimension is C, number of labels. 

theta : :py:class:`numpy.ndarray` 

Log of class conditional probabilities, whose dimension is C-by-D, 

where D is number of features. 

Examples 

-------- 

>>> from pyspark.mllib.linalg import SparseVector 

>>> data = [ 

... LabeledPoint(0.0, [0.0, 0.0]), 

... LabeledPoint(0.0, [0.0, 1.0]), 

... LabeledPoint(1.0, [1.0, 0.0]), 

... ] 

>>> model = NaiveBayes.train(sc.parallelize(data)) 

>>> model.predict(numpy.array([0.0, 1.0])) 

0.0 

>>> model.predict(numpy.array([1.0, 0.0])) 

1.0 

>>> model.predict(sc.parallelize([[1.0, 0.0]])).collect() 

[1.0] 

>>> sparse_data = [ 

... LabeledPoint(0.0, SparseVector(2, {1: 0.0})), 

... LabeledPoint(0.0, SparseVector(2, {1: 1.0})), 

... LabeledPoint(1.0, SparseVector(2, {0: 1.0})) 

... ] 

>>> model = NaiveBayes.train(sc.parallelize(sparse_data)) 

>>> model.predict(SparseVector(2, {1: 1.0})) 

0.0 

>>> model.predict(SparseVector(2, {0: 1.0})) 

1.0 

>>> import os, tempfile 

>>> path = tempfile.mkdtemp() 

>>> model.save(sc, path) 

>>> sameModel = NaiveBayesModel.load(sc, path) 

>>> sameModel.predict(SparseVector(2, {0: 1.0})) == model.predict(SparseVector(2, {0: 1.0})) 

True 

>>> from shutil import rmtree 

>>> try: 

... rmtree(path) 

... except OSError: 

... pass 

""" 

def __init__(self, labels, pi, theta): 

self.labels = labels 

self.pi = pi 

self.theta = theta 

@since('0.9.0') 

def predict(self, x): 

""" 

Return the most likely class for a data vector 

or an RDD of vectors 

""" 

if isinstance(x, RDD): 

return x.map(lambda v: self.predict(v)) 

x = _convert_to_vector(x) 

return self.labels[numpy.argmax(self.pi + x.dot(self.theta.transpose()))] 

def save(self, sc, path): 

""" 

Save this model to the given path. 

""" 

java_labels = _py2java(sc, self.labels.tolist()) 

java_pi = _py2java(sc, self.pi.tolist()) 

java_theta = _py2java(sc, self.theta.tolist()) 

java_model = sc._jvm.org.apache.spark.mllib.classification.NaiveBayesModel( 

java_labels, java_pi, java_theta) 

java_model.save(sc._jsc.sc(), path) 

@classmethod 

@since('1.4.0') 

def load(cls, sc, path): 

""" 

Load a model from the given path. 

""" 

java_model = sc._jvm.org.apache.spark.mllib.classification.NaiveBayesModel.load( 

sc._jsc.sc(), path) 

# Can not unpickle array.array from Pyrolite in Python3 with "bytes" 

py_labels = _java2py(sc, java_model.labels(), "latin1") 

py_pi = _java2py(sc, java_model.pi(), "latin1") 

py_theta = _java2py(sc, java_model.theta(), "latin1") 

return NaiveBayesModel(py_labels, py_pi, numpy.array(py_theta)) 

class NaiveBayes(object): 

""" 

Train a Multinomial Naive Bayes model. 

.. versionadded:: 0.9.0 

""" 

@classmethod 

def train(cls, data, lambda_=1.0): 

""" 

Train a Naive Bayes model given an RDD of (label, features) 

vectors. 

This is the `Multinomial NB <http://tinyurl.com/lsdw6p>`_ which 

can handle all kinds of discrete data. For example, by 

converting documents into TF-IDF vectors, it can be used for 

document classification. By making every vector a 0-1 vector, 

it can also be used as `Bernoulli NB <http://tinyurl.com/p7c96j6>`_. 

The input feature values must be nonnegative. 

.. versionadded:: 0.9.0 

Parameters 

---------- 

data : :py:class:`pyspark.RDD` 

The training data, an RDD of :py:class:`pyspark.mllib.regression.LabeledPoint`. 

lambda\\_ : float, optional 

The smoothing parameter. 

(default: 1.0) 

""" 

first = data.first() 

718 ↛ 719line 718 didn't jump to line 719, because the condition on line 718 was never true if not isinstance(first, LabeledPoint): 

raise ValueError("`data` should be an RDD of LabeledPoint") 

labels, pi, theta = callMLlibFunc("trainNaiveBayesModel", data, lambda_) 

return NaiveBayesModel(labels.toArray(), pi.toArray(), numpy.array(theta)) 

@inherit_doc 

class StreamingLogisticRegressionWithSGD(StreamingLinearAlgorithm): 

""" 

Train or predict a logistic regression model on streaming data. 

Training uses Stochastic Gradient Descent to update the model based on 

each new batch of incoming data from a DStream. 

Each batch of data is assumed to be an RDD of LabeledPoints. 

The number of data points per batch can vary, but the number 

of features must be constant. An initial weight 

vector must be provided. 

.. versionadded:: 1.5.0 

Parameters 

---------- 

stepSize : float, optional 

Step size for each iteration of gradient descent. 

(default: 0.1) 

numIterations : int, optional 

Number of iterations run for each batch of data. 

(default: 50) 

miniBatchFraction : float, optional 

Fraction of each batch of data to use for updates. 

(default: 1.0) 

regParam : float, optional 

L2 Regularization parameter. 

(default: 0.0) 

convergenceTol : float, optional 

Value used to determine when to terminate iterations. 

(default: 0.001) 

""" 

def __init__(self, stepSize=0.1, numIterations=50, miniBatchFraction=1.0, regParam=0.0, 

convergenceTol=0.001): 

self.stepSize = stepSize 

self.numIterations = numIterations 

self.regParam = regParam 

self.miniBatchFraction = miniBatchFraction 

self.convergenceTol = convergenceTol 

self._model = None 

super(StreamingLogisticRegressionWithSGD, self).__init__( 

model=self._model) 

@since('1.5.0') 

def setInitialWeights(self, initialWeights): 

""" 

Set the initial value of weights. 

This must be set before running trainOn and predictOn. 

""" 

initialWeights = _convert_to_vector(initialWeights) 

# LogisticRegressionWithSGD does only binary classification. 

self._model = LogisticRegressionModel( 

initialWeights, 0, initialWeights.size, 2) 

return self 

@since('1.5.0') 

def trainOn(self, dstream): 

"""Train the model on the incoming dstream.""" 

self._validate(dstream) 

def update(rdd): 

# LogisticRegressionWithSGD.train raises an error for an empty RDD. 

if not rdd.isEmpty(): 

self._model = LogisticRegressionWithSGD.train( 

rdd, self.numIterations, self.stepSize, 

self.miniBatchFraction, self._model.weights, 

regParam=self.regParam, convergenceTol=self.convergenceTol) 

dstream.foreachRDD(update) 

def _test(): 

import doctest 

from pyspark.sql import SparkSession 

import pyspark.mllib.classification 

globs = pyspark.mllib.classification.__dict__.copy() 

spark = SparkSession.builder\ 

.master("local[4]")\ 

.appName("mllib.classification tests")\ 

.getOrCreate() 

globs['sc'] = spark.sparkContext 

(failure_count, test_count) = doctest.testmod(globs=globs, optionflags=doctest.ELLIPSIS) 

spark.stop() 

809 ↛ 810line 809 didn't jump to line 810, because the condition on line 809 was never true if failure_count: 

sys.exit(-1) 

if __name__ == "__main__": 

_test()