1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
|
/*
* 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.
*/
package org.apache.spark.ml.classification
import org.apache.spark.annotation.AlphaComponent
import org.apache.spark.ml.param._
import org.apache.spark.mllib.classification.LogisticRegressionWithLBFGS
import org.apache.spark.mllib.linalg.{VectorUDT, BLAS, Vector, Vectors}
import org.apache.spark.sql.DataFrame
import org.apache.spark.sql.functions._
import org.apache.spark.sql.types.DoubleType
import org.apache.spark.storage.StorageLevel
/**
* Params for logistic regression.
*/
private[classification] trait LogisticRegressionParams extends ProbabilisticClassifierParams
with HasRegParam with HasMaxIter with HasFitIntercept with HasThreshold
/**
* :: AlphaComponent ::
*
* Logistic regression.
* Currently, this class only supports binary classification.
*/
@AlphaComponent
class LogisticRegression
extends ProbabilisticClassifier[Vector, LogisticRegression, LogisticRegressionModel]
with LogisticRegressionParams {
setRegParam(0.1)
setMaxIter(100)
setThreshold(0.5)
/** @group setParam */
def setRegParam(value: Double): this.type = set(regParam, value)
/** @group setParam */
def setMaxIter(value: Int): this.type = set(maxIter, value)
/** @group setParam */
def setFitIntercept(value: Boolean): this.type = set(fitIntercept, value)
/** @group setParam */
def setThreshold(value: Double): this.type = set(threshold, value)
override protected def train(dataset: DataFrame, paramMap: ParamMap): LogisticRegressionModel = {
// Extract columns from data. If dataset is persisted, do not persist oldDataset.
val oldDataset = extractLabeledPoints(dataset, paramMap)
val handlePersistence = dataset.rdd.getStorageLevel == StorageLevel.NONE
if (handlePersistence) {
oldDataset.persist(StorageLevel.MEMORY_AND_DISK)
}
// Train model
val lr = new LogisticRegressionWithLBFGS()
.setIntercept(paramMap(fitIntercept))
lr.optimizer
.setRegParam(paramMap(regParam))
.setNumIterations(paramMap(maxIter))
val oldModel = lr.run(oldDataset)
val lrm = new LogisticRegressionModel(this, paramMap, oldModel.weights, oldModel.intercept)
if (handlePersistence) {
oldDataset.unpersist()
}
lrm
}
}
/**
* :: AlphaComponent ::
*
* Model produced by [[LogisticRegression]].
*/
@AlphaComponent
class LogisticRegressionModel private[ml] (
override val parent: LogisticRegression,
override val fittingParamMap: ParamMap,
val weights: Vector,
val intercept: Double)
extends ProbabilisticClassificationModel[Vector, LogisticRegressionModel]
with LogisticRegressionParams {
setThreshold(0.5)
/** @group setParam */
def setThreshold(value: Double): this.type = set(threshold, value)
private val margin: Vector => Double = (features) => {
BLAS.dot(features, weights) + intercept
}
private val score: Vector => Double = (features) => {
val m = margin(features)
1.0 / (1.0 + math.exp(-m))
}
override def transform(dataset: DataFrame, paramMap: ParamMap): DataFrame = {
// This is overridden (a) to be more efficient (avoiding re-computing values when creating
// multiple output columns) and (b) to handle threshold, which the abstractions do not use.
// TODO: We should abstract away the steps defined by UDFs below so that the abstractions
// can call whichever UDFs are needed to create the output columns.
// Check schema
transformSchema(dataset.schema, paramMap, logging = true)
val map = this.paramMap ++ paramMap
// Output selected columns only.
// This is a bit complicated since it tries to avoid repeated computation.
// rawPrediction (-margin, margin)
// probability (1.0-score, score)
// prediction (max margin)
var tmpData = dataset
var numColsOutput = 0
if (map(rawPredictionCol) != "") {
val features2raw: Vector => Vector = (features) => predictRaw(features)
tmpData = tmpData.withColumn(map(rawPredictionCol),
callUDF(features2raw, new VectorUDT, col(map(featuresCol))))
numColsOutput += 1
}
if (map(probabilityCol) != "") {
if (map(rawPredictionCol) != "") {
val raw2prob = udf { (rawPreds: Vector) =>
val prob1 = 1.0 / (1.0 + math.exp(-rawPreds(1)))
Vectors.dense(1.0 - prob1, prob1): Vector
}
tmpData = tmpData.withColumn(map(probabilityCol), raw2prob(col(map(rawPredictionCol))))
} else {
val features2prob = udf { (features: Vector) => predictProbabilities(features) : Vector }
tmpData = tmpData.withColumn(map(probabilityCol), features2prob(col(map(featuresCol))))
}
numColsOutput += 1
}
if (map(predictionCol) != "") {
val t = map(threshold)
if (map(probabilityCol) != "") {
val predict = udf { probs: Vector =>
if (probs(1) > t) 1.0 else 0.0
}
tmpData = tmpData.withColumn(map(predictionCol), predict(col(map(probabilityCol))))
} else if (map(rawPredictionCol) != "") {
val predict = udf { rawPreds: Vector =>
val prob1 = 1.0 / (1.0 + math.exp(-rawPreds(1)))
if (prob1 > t) 1.0 else 0.0
}
tmpData = tmpData.withColumn(map(predictionCol), predict(col(map(rawPredictionCol))))
} else {
val predict = udf { features: Vector => this.predict(features) }
tmpData = tmpData.withColumn(map(predictionCol), predict(col(map(featuresCol))))
}
numColsOutput += 1
}
if (numColsOutput == 0) {
this.logWarning(s"$uid: LogisticRegressionModel.transform() was called as NOOP" +
" since no output columns were set.")
}
tmpData
}
override val numClasses: Int = 2
/**
* Predict label for the given feature vector.
* The behavior of this can be adjusted using [[threshold]].
*/
override protected def predict(features: Vector): Double = {
if (score(features) > paramMap(threshold)) 1 else 0
}
override protected def predictProbabilities(features: Vector): Vector = {
val s = score(features)
Vectors.dense(1.0 - s, s)
}
override protected def predictRaw(features: Vector): Vector = {
val m = margin(features)
Vectors.dense(0.0, m)
}
override protected def copy(): LogisticRegressionModel = {
val m = new LogisticRegressionModel(parent, fittingParamMap, weights, intercept)
Params.inheritValues(this.paramMap, this, m)
m
}
}
|
