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
|
/*
* 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
import org.apache.spark.annotation.{DeveloperApi, Since}
import org.apache.spark.ml.param._
import org.apache.spark.ml.param.shared._
import org.apache.spark.ml.util.SchemaUtils
import org.apache.spark.mllib.linalg.{Vector, VectorUDT}
import org.apache.spark.mllib.regression.LabeledPoint
import org.apache.spark.rdd.RDD
import org.apache.spark.sql.{DataFrame, Dataset, Row}
import org.apache.spark.sql.functions._
import org.apache.spark.sql.types.{DataType, DoubleType, StructType}
/**
* (private[ml]) Trait for parameters for prediction (regression and classification).
*/
private[ml] trait PredictorParams extends Params
with HasLabelCol with HasFeaturesCol with HasPredictionCol {
/**
* Validates and transforms the input schema with the provided param map.
*
* @param schema input schema
* @param fitting whether this is in fitting
* @param featuresDataType SQL DataType for FeaturesType.
* E.g., [[org.apache.spark.mllib.linalg.VectorUDT]] for vector features.
* @return output schema
*/
protected def validateAndTransformSchema(
schema: StructType,
fitting: Boolean,
featuresDataType: DataType): StructType = {
// TODO: Support casting Array[Double] and Array[Float] to Vector when FeaturesType = Vector
SchemaUtils.checkColumnType(schema, $(featuresCol), featuresDataType)
if (fitting) {
SchemaUtils.checkNumericType(schema, $(labelCol))
}
SchemaUtils.appendColumn(schema, $(predictionCol), DoubleType)
}
}
/**
* :: DeveloperApi ::
* Abstraction for prediction problems (regression and classification).
*
* @tparam FeaturesType Type of features.
* E.g., [[org.apache.spark.mllib.linalg.VectorUDT]] for vector features.
* @tparam Learner Specialization of this class. If you subclass this type, use this type
* parameter to specify the concrete type.
* @tparam M Specialization of [[PredictionModel]]. If you subclass this type, use this type
* parameter to specify the concrete type for the corresponding model.
*/
@DeveloperApi
abstract class Predictor[
FeaturesType,
Learner <: Predictor[FeaturesType, Learner, M],
M <: PredictionModel[FeaturesType, M]]
extends Estimator[M] with PredictorParams {
/** @group setParam */
def setLabelCol(value: String): Learner = set(labelCol, value).asInstanceOf[Learner]
/** @group setParam */
def setFeaturesCol(value: String): Learner = set(featuresCol, value).asInstanceOf[Learner]
/** @group setParam */
def setPredictionCol(value: String): Learner = set(predictionCol, value).asInstanceOf[Learner]
override def fit(dataset: Dataset[_]): M = {
// This handles a few items such as schema validation.
// Developers only need to implement train().
transformSchema(dataset.schema, logging = true)
copyValues(train(dataset).setParent(this))
}
override def copy(extra: ParamMap): Learner
/**
* Train a model using the given dataset and parameters.
* Developers can implement this instead of [[fit()]] to avoid dealing with schema validation
* and copying parameters into the model.
*
* @param dataset Training dataset
* @return Fitted model
*/
protected def train(dataset: Dataset[_]): M
/**
* Returns the SQL DataType corresponding to the FeaturesType type parameter.
*
* This is used by [[validateAndTransformSchema()]].
* This workaround is needed since SQL has different APIs for Scala and Java.
*
* The default value is VectorUDT, but it may be overridden if FeaturesType is not Vector.
*/
private[ml] def featuresDataType: DataType = new VectorUDT
override def transformSchema(schema: StructType): StructType = {
validateAndTransformSchema(schema, fitting = true, featuresDataType)
}
/**
* Extract [[labelCol]] and [[featuresCol]] from the given dataset,
* and put it in an RDD with strong types.
*/
protected def extractLabeledPoints(dataset: Dataset[_]): RDD[LabeledPoint] = {
dataset.select(col($(labelCol)).cast(DoubleType), col($(featuresCol))).rdd.map {
case Row(label: Double, features: Vector) => LabeledPoint(label, features)
}
}
}
/**
* :: DeveloperApi ::
* Abstraction for a model for prediction tasks (regression and classification).
*
* @tparam FeaturesType Type of features.
* E.g., [[org.apache.spark.mllib.linalg.VectorUDT]] for vector features.
* @tparam M Specialization of [[PredictionModel]]. If you subclass this type, use this type
* parameter to specify the concrete type for the corresponding model.
*/
@DeveloperApi
abstract class PredictionModel[FeaturesType, M <: PredictionModel[FeaturesType, M]]
extends Model[M] with PredictorParams {
/** @group setParam */
def setFeaturesCol(value: String): M = set(featuresCol, value).asInstanceOf[M]
/** @group setParam */
def setPredictionCol(value: String): M = set(predictionCol, value).asInstanceOf[M]
/** Returns the number of features the model was trained on. If unknown, returns -1 */
@Since("1.6.0")
def numFeatures: Int = -1
/**
* Returns the SQL DataType corresponding to the FeaturesType type parameter.
*
* This is used by [[validateAndTransformSchema()]].
* This workaround is needed since SQL has different APIs for Scala and Java.
*
* The default value is VectorUDT, but it may be overridden if FeaturesType is not Vector.
*/
protected def featuresDataType: DataType = new VectorUDT
override def transformSchema(schema: StructType): StructType = {
validateAndTransformSchema(schema, fitting = false, featuresDataType)
}
/**
* Transforms dataset by reading from [[featuresCol]], calling [[predict()]], and storing
* the predictions as a new column [[predictionCol]].
*
* @param dataset input dataset
* @return transformed dataset with [[predictionCol]] of type [[Double]]
*/
override def transform(dataset: Dataset[_]): DataFrame = {
transformSchema(dataset.schema, logging = true)
if ($(predictionCol).nonEmpty) {
transformImpl(dataset)
} else {
this.logWarning(s"$uid: Predictor.transform() was called as NOOP" +
" since no output columns were set.")
dataset.toDF
}
}
protected def transformImpl(dataset: Dataset[_]): DataFrame = {
val predictUDF = udf { (features: Any) =>
predict(features.asInstanceOf[FeaturesType])
}
dataset.withColumn($(predictionCol), predictUDF(col($(featuresCol))))
}
/**
* Predict label for the given features.
* This internal method is used to implement [[transform()]] and output [[predictionCol]].
*/
protected def predict(features: FeaturesType): Double
}
|
