Github user hhbyyh commented on a diff in the pull request:
https://github.com/apache/spark/pull/15415#discussion_r102840479
--- Diff: mllib/src/main/scala/org/apache/spark/ml/fpm/FPGrowth.scala ---
@@ -0,0 +1,346 @@
+/*
+ * 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.fpm
+
+import scala.collection.mutable.ArrayBuffer
+import scala.reflect.ClassTag
+
+import org.apache.hadoop.fs.Path
+
+import org.apache.spark.annotation.{Experimental, Since}
+import org.apache.spark.ml.{Estimator, Model}
+import org.apache.spark.ml.param._
+import org.apache.spark.ml.param.shared.{HasFeaturesCol, HasPredictionCol}
+import org.apache.spark.ml.util._
+import org.apache.spark.mllib.fpm.{AssociationRules =>
MLlibAssociationRules,
+ FPGrowth => MLlibFPGrowth}
+import org.apache.spark.mllib.fpm.FPGrowth.FreqItemset
+import org.apache.spark.sql._
+import org.apache.spark.sql.functions._
+import org.apache.spark.sql.types._
+
+/**
+ * Common params for FPGrowth and FPGrowthModel
+ */
+private[fpm] trait FPGrowthParams extends Params with HasFeaturesCol with
HasPredictionCol {
+
+ /**
+ * Minimal support level of the frequent pattern. [0.0, 1.0]. Any
pattern that appears
+ * more than (minSupport * size-of-the-dataset) times will be output
+ * Default: 0.3
+ * @group param
+ */
+ @Since("2.2.0")
+ val minSupport: DoubleParam = new DoubleParam(this, "minSupport",
+ "the minimal support level of a frequent pattern",
+ ParamValidators.inRange(0.0, 1.0))
+ setDefault(minSupport -> 0.3)
+
+ /** @group getParam */
+ @Since("2.2.0")
+ def getMinSupport: Double = $(minSupport)
+
+ /**
+ * Number of partitions (>=1) used by parallel FP-growth. By default the
param is not set, and
+ * partition number of the input dataset is used.
+ * @group expertParam
+ */
+ @Since("2.2.0")
+ val numPartitions: IntParam = new IntParam(this, "numPartitions",
+ "Number of partitions used by parallel FP-growth",
ParamValidators.gtEq[Int](1))
+
+ /** @group expertGetParam */
+ @Since("2.2.0")
+ def getNumPartitions: Int = $(numPartitions)
+
+ /**
+ * Minimal confidence for generating Association Rule.
+ * Note that minConfidence has no effect during fitting.
+ * Default: 0.8
+ * @group param
+ */
+ @Since("2.2.0")
+ val minConfidence: DoubleParam = new DoubleParam(this, "minConfidence",
+ "minimal confidence for generating Association Rule",
+ ParamValidators.inRange(0.0, 1.0))
+ setDefault(minConfidence -> 0.8)
+
+ /** @group getParam */
+ @Since("2.2.0")
+ def getMinConfidence: Double = $(minConfidence)
+
+ /**
+ * Validates and transforms the input schema.
+ * @param schema input schema
+ * @return output schema
+ */
+ @Since("2.2.0")
+ protected def validateAndTransformSchema(schema: StructType): StructType
= {
+ val inputType = schema($(featuresCol)).dataType
+ require(inputType.isInstanceOf[ArrayType],
+ s"The input column must be ArrayType, but got $inputType.")
+ SchemaUtils.appendColumn(schema, $(predictionCol),
schema($(featuresCol)).dataType)
+ }
+}
+
+/**
+ * :: Experimental ::
+ * A parallel FP-growth algorithm to mine frequent itemsets. The algorithm
is described in
+ * <a href="http://dx.doi.org/10.1145/1454008.1454027";>Li et al., PFP:
Parallel FP-Growth for Query
+ * Recommendation</a>. PFP distributes computation in such a way that each
worker executes an
+ * independent group of mining tasks. The FP-Growth algorithm is described
in
+ * <a href="http://dx.doi.org/10.1145/335191.335372";>Han et al., Mining
frequent patterns without
+ * candidate generation</a>.
+ *
+ * @see <a href="http://en.wikipedia.org/wiki/Association_rule_learning";>
+ * Association rule learning (Wikipedia)</a>
+ */
+@Since("2.2.0")
+@Experimental
+class FPGrowth @Since("2.2.0") (
+ @Since("2.2.0") override val uid: String)
+ extends Estimator[FPGrowthModel] with FPGrowthParams with
DefaultParamsWritable {
+
+ @Since("2.2.0")
+ def this() = this(Identifiable.randomUID("fpgrowth"))
+
+ /** @group setParam */
+ @Since("2.2.0")
+ def setMinSupport(value: Double): this.type = set(minSupport, value)
+
+ /** @group expertSetParam */
+ @Since("2.2.0")
+ def setNumPartitions(value: Int): this.type = set(numPartitions, value)
+
+ /** @group setParam */
+ @Since("2.2.0")
+ def setMinConfidence(value: Double): this.type = set(minConfidence,
value)
+
+ /** @group setParam */
+ @Since("2.2.0")
+ def setFeaturesCol(value: String): this.type = set(featuresCol, value)
+
+ /** @group setParam */
+ @Since("2.2.0")
+ def setPredictionCol(value: String): this.type = set(predictionCol,
value)
+
+ @Since("2.2.0")
+ override def fit(dataset: Dataset[_]): FPGrowthModel = {
+ transformSchema(dataset.schema, logging = true)
+ genericFit(dataset)
+ }
+
+ private def genericFit[T: ClassTag](dataset: Dataset[_]): FPGrowthModel
= {
+ val data = dataset.select($(featuresCol))
+ val items = data.where(col($(featuresCol)).isNotNull).rdd.map(r =>
r.getSeq[T](0).toArray)
+ val mllibFP = new MLlibFPGrowth().setMinSupport($(minSupport))
+ if (isSet(numPartitions)) {
+ mllibFP.setNumPartitions($(numPartitions))
+ }
+ val parentModel = mllibFP.run(items)
+ val rows = parentModel.freqItemsets.map(f => Row(f.items, f.freq))
+
+ val schema = StructType(Seq(
+ StructField("items", dataset.schema($(featuresCol)).dataType,
nullable = false),
+ StructField("freq", LongType, nullable = false)))
+ val frequentItems = dataset.sparkSession.createDataFrame(rows, schema)
+ copyValues(new FPGrowthModel(uid, frequentItems)).setParent(this)
+ }
+
+ @Since("2.2.0")
+ override def transformSchema(schema: StructType): StructType = {
+ validateAndTransformSchema(schema)
+ }
+
+ @Since("2.2.0")
+ override def copy(extra: ParamMap): FPGrowth = defaultCopy(extra)
+}
+
+
+@Since("2.2.0")
+object FPGrowth extends DefaultParamsReadable[FPGrowth] {
+
+ @Since("2.2.0")
+ override def load(path: String): FPGrowth = super.load(path)
+}
+
+/**
+ * :: Experimental ::
+ * Model fitted by FPGrowth.
+ *
+ * @param freqItemsets frequent items in the format of
DataFrame("items"[Seq], "freq"[Long])
+ */
+@Since("2.2.0")
+@Experimental
+class FPGrowthModel private[ml] (
+ @Since("2.2.0") override val uid: String,
+ @transient val freqItemsets: DataFrame)
+ extends Model[FPGrowthModel] with FPGrowthParams with MLWritable {
+
+ /** @group setParam */
+ @Since("2.2.0")
+ def setMinConfidence(value: Double): this.type = set(minConfidence,
value)
+
+ /** @group setParam */
+ @Since("2.2.0")
+ def setFeaturesCol(value: String): this.type = set(featuresCol, value)
+
+ /** @group setParam */
+ @Since("2.2.0")
+ def setPredictionCol(value: String): this.type = set(predictionCol,
value)
+
+ /**
+ * Get association rules fitted by AssociationRules using the
minConfidence. Returns a dataframe
+ * with three fields, "antecedent", "consequent" and "confidence", where
"antecedent" and
+ * "consequent" are Array[T] and "confidence" is Double.
+ */
+ @Since("2.2.0")
+ @transient lazy val associationRules: DataFrame = {
+ val freqItems = freqItemsets
+ AssociationRules.getAssociationRulesFromFP(freqItems, "items", "freq",
$(minConfidence))
+ }
+
+ /**
+ * The transform method first generates the association rules according
to the frequent itemsets.
+ * Then for each association rule, it will examine the input items
against antecedents and
+ * summarize the consequents as prediction. The prediction column has
the same data type as the
+ * input column. (Array[T])
+ * Note that internally it uses Cartesian join and may exhaust memory
for large datasets.
+ */
+ @Since("2.2.0")
+ override def transform(dataset: Dataset[_]): DataFrame = {
+ transformSchema(dataset.schema, logging = true)
+ genericTransform(dataset)
+ }
+
+ private def genericTransform[T](dataset: Dataset[_]): DataFrame = {
+ // use index to perform the join and aggregateByKey, and keep the
original order after join.
+ val indexToItems = dataset.select($(featuresCol)).rdd.map(r =>
r.getSeq[T](0))
+ .zipWithIndex().map(_.swap)
+ val rulesRDD = associationRules.select("antecedent", "consequent").rdd
+ .map(r => (r.getSeq[T](0), r.getSeq[T](1)))
+
+ val indexToConsequents = indexToItems.cartesian(rulesRDD).map {
+ case ((id, items), (antecedent, consequent)) =>
+ val consequents = if (items != null) {
+ val itemSet = items.toSet
+ if (antecedent.forall(itemSet.contains)) {
+ consequent.filterNot(itemSet.contains)
+ } else {
+ Seq.empty
+ }
+ } else {
+ Seq.empty
+ }
+ (id, consequents)
+ }.aggregateByKey(new ArrayBuffer[T])((ar, seq) => ar ++= seq, (ar,
seq) => ar ++= seq)
+ .map { case (index, cons) => (index, cons.distinct) }
+
+ val rowAndConsequents = dataset.toDF().rdd.zipWithIndex().map(_.swap)
+ .join(indexToConsequents).sortByKey(ascending = true,
dataset.rdd.getNumPartitions)
--- End diff --
Shall we try to keep the original order of the input dataset? The time cost
is about 5% of the total transform time.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
