pyspark.mllib.regression

1 # 2 # Licensed to the Apache Software Foundation (ASF) under one or more 3 # contributor license agreements. See the NOTICE file distributed with 4 # this work for additional information regarding copyright ownership. 5 # The ASF licenses this file to You under the Apache License, Version 2.0 6 # (the "License"); you may not use this file except in compliance with 7 # the License. You may obtain a copy of the License at 8 # 9 # http://www.apache.org/licenses/LICENSE-2.0 10 # 11 # Unless required by applicable law or agreed to in writing, software 12 # distributed under the License is distributed on an "AS IS" BASIS, 13 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 # See the License for the specific language governing permissions and 15 # limitations under the License. 16 # 17 18 from numpy import array, ndarray 19 from pyspark import SparkContext 20 from pyspark.mllib._common import \ 21 _dot, _get_unmangled_rdd, _get_unmangled_double_vector_rdd, \ 22 _serialize_double_matrix, _deserialize_double_matrix, \ 23 _serialize_double_vector, _deserialize_double_vector, \ 24 _get_initial_weights, _serialize_rating, _regression_train_wrapper, \ 25 _linear_predictor_typecheck, _have_scipy, _scipy_issparse 26 from pyspark.mllib.linalg import SparseVector

27 28 29 -class LabeledPoint(object):

30 """ 31 The features and labels of a data point. 32 33 @param label: Label for this data point. 34 @param features: Vector of features for this point (NumPy array, list, 35 pyspark.mllib.linalg.SparseVector, or scipy.sparse column matrix) 36 """

37 - def __init__(self, label, features):

38 self.label = label 39 if (type(features) == ndarray or type(features) == SparseVector 40 or (_have_scipy and _scipy_issparse(features))): 41 self.features = features 42 elif type(features) == list: 43 self.features = array(features) 44 else: 45 raise TypeError("Expected NumPy array, list, SparseVector, or scipy.sparse matrix")

46

47 48 -class LinearModel(object):

49 """A linear model that has a vector of coefficients and an intercept."""

50 - def __init__(self, weights, intercept):

51 self._coeff = weights 52 self._intercept = intercept

53 54 @property

55 - def weights(self):

56 return self._coeff

57 58 @property

59 - def intercept(self):

60 return self._intercept

61

62 63 -class LinearRegressionModelBase(LinearModel):

64 """A linear regression model. 65 66 >>> lrmb = LinearRegressionModelBase(array([1.0, 2.0]), 0.1) 67 >>> abs(lrmb.predict(array([-1.03, 7.777])) - 14.624) < 1e-6 68 True 69 >>> abs(lrmb.predict(SparseVector(2, {0: -1.03, 1: 7.777})) - 14.624) < 1e-6 70 True 71 """

72 - def predict(self, x):

73 """Predict the value of the dependent variable given a vector x""" 74 """containing values for the independent variables.""" 75 _linear_predictor_typecheck(x, self._coeff) 76 return _dot(x, self._coeff) + self._intercept

77

78 79 -class LinearRegressionModel(LinearRegressionModelBase):

80 """A linear regression model derived from a least-squares fit. 81 82 >>> from pyspark.mllib.regression import LabeledPoint 83 >>> data = [ 84 ... LabeledPoint(0.0, [0.0]), 85 ... LabeledPoint(1.0, [1.0]), 86 ... LabeledPoint(3.0, [2.0]), 87 ... LabeledPoint(2.0, [3.0]) 88 ... ] 89 >>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), initialWeights=array([1.0])) 90 >>> abs(lrm.predict(array([0.0])) - 0) < 0.5 91 True 92 >>> abs(lrm.predict(array([1.0])) - 1) < 0.5 93 True 94 >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 95 True 96 >>> data = [ 97 ... LabeledPoint(0.0, SparseVector(1, {0: 0.0})), 98 ... LabeledPoint(1.0, SparseVector(1, {0: 1.0})), 99 ... LabeledPoint(3.0, SparseVector(1, {0: 2.0})), 100 ... LabeledPoint(2.0, SparseVector(1, {0: 3.0})) 101 ... ] 102 >>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), initialWeights=array([1.0])) 103 >>> abs(lrm.predict(array([0.0])) - 0) < 0.5 104 True 105 >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 106 True 107 """

108

109 110 -class LinearRegressionWithSGD(object):

111 @classmethod

112 - def train(cls, data, iterations=100, step=1.0, 113 miniBatchFraction=1.0, initialWeights=None):

114 """Train a linear regression model on the given data.""" 115 sc = data.context 116 train_f = lambda d, i: sc._jvm.PythonMLLibAPI().trainLinearRegressionModelWithSGD( 117 d._jrdd, iterations, step, miniBatchFraction, i) 118 return _regression_train_wrapper(sc, train_f, LinearRegressionModel, data, initialWeights)

119

120 121 -class LassoModel(LinearRegressionModelBase):

122 """A linear regression model derived from a least-squares fit with an 123 l_1 penalty term. 124 125 >>> from pyspark.mllib.regression import LabeledPoint 126 >>> data = [ 127 ... LabeledPoint(0.0, [0.0]), 128 ... LabeledPoint(1.0, [1.0]), 129 ... LabeledPoint(3.0, [2.0]), 130 ... LabeledPoint(2.0, [3.0]) 131 ... ] 132 >>> lrm = LassoWithSGD.train(sc.parallelize(data), initialWeights=array([1.0])) 133 >>> abs(lrm.predict(array([0.0])) - 0) < 0.5 134 True 135 >>> abs(lrm.predict(array([1.0])) - 1) < 0.5 136 True 137 >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 138 True 139 >>> data = [ 140 ... LabeledPoint(0.0, SparseVector(1, {0: 0.0})), 141 ... LabeledPoint(1.0, SparseVector(1, {0: 1.0})), 142 ... LabeledPoint(3.0, SparseVector(1, {0: 2.0})), 143 ... LabeledPoint(2.0, SparseVector(1, {0: 3.0})) 144 ... ] 145 >>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), initialWeights=array([1.0])) 146 >>> abs(lrm.predict(array([0.0])) - 0) < 0.5 147 True 148 >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 149 True 150 """

151

152 153 -class LassoWithSGD(object):

154 @classmethod

155 - def train(cls, data, iterations=100, step=1.0, regParam=1.0, 156 miniBatchFraction=1.0, initialWeights=None):

157 """Train a Lasso regression model on the given data.""" 158 sc = data.context 159 train_f = lambda d, i: sc._jvm.PythonMLLibAPI().trainLassoModelWithSGD( 160 d._jrdd, iterations, step, regParam, miniBatchFraction, i) 161 return _regression_train_wrapper(sc, train_f, LassoModel, data, initialWeights)

162

163 164 -class RidgeRegressionModel(LinearRegressionModelBase):

165 """A linear regression model derived from a least-squares fit with an 166 l_2 penalty term. 167 168 >>> from pyspark.mllib.regression import LabeledPoint 169 >>> data = [ 170 ... LabeledPoint(0.0, [0.0]), 171 ... LabeledPoint(1.0, [1.0]), 172 ... LabeledPoint(3.0, [2.0]), 173 ... LabeledPoint(2.0, [3.0]) 174 ... ] 175 >>> lrm = RidgeRegressionWithSGD.train(sc.parallelize(data), initialWeights=array([1.0])) 176 >>> abs(lrm.predict(array([0.0])) - 0) < 0.5 177 True 178 >>> abs(lrm.predict(array([1.0])) - 1) < 0.5 179 True 180 >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 181 True 182 >>> data = [ 183 ... LabeledPoint(0.0, SparseVector(1, {0: 0.0})), 184 ... LabeledPoint(1.0, SparseVector(1, {0: 1.0})), 185 ... LabeledPoint(3.0, SparseVector(1, {0: 2.0})), 186 ... LabeledPoint(2.0, SparseVector(1, {0: 3.0})) 187 ... ] 188 >>> lrm = LinearRegressionWithSGD.train(sc.parallelize(data), initialWeights=array([1.0])) 189 >>> abs(lrm.predict(array([0.0])) - 0) < 0.5 190 True 191 >>> abs(lrm.predict(SparseVector(1, {0: 1.0})) - 1) < 0.5 192 True 193 """

194

195 196 -class RidgeRegressionWithSGD(object):

197 @classmethod

198 - def train(cls, data, iterations=100, step=1.0, regParam=1.0, 199 miniBatchFraction=1.0, initialWeights=None):

200 """Train a ridge regression model on the given data.""" 201 sc = data.context 202 train_func = lambda d, i: sc._jvm.PythonMLLibAPI().trainRidgeModelWithSGD( 203 d._jrdd, iterations, step, regParam, miniBatchFraction, i) 204 return _regression_train_wrapper(sc, train_func, RidgeRegressionModel, data, initialWeights)

205

206 207 -def _test():

208 import doctest 209 globs = globals().copy() 210 globs['sc'] = SparkContext('local[4]', 'PythonTest', batchSize=2) 211 (failure_count, test_count) = doctest.testmod(globs=globs, optionflags=doctest.ELLIPSIS) 212 globs['sc'].stop() 213 if failure_count: 214 exit(-1)

215 216 if __name__ == "__main__": 217 _test() 218

Source Code for Module pyspark.mllib.regression