Github user srowen commented on a diff in the pull request:
https://github.com/apache/spark/pull/12461#discussion_r60534110
--- Diff:
mllib/src/main/scala/org/apache/spark/ml/evaluation/RankingEvaluator.scala ---
@@ -0,0 +1,287 @@
+/*
+ * 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.evaluation
+
+import scala.reflect.ClassTag
+
+import org.apache.spark.SparkContext
+import org.apache.spark.annotation.{Experimental, Since}
+import org.apache.spark.internal.Logging
+import org.apache.spark.ml.param.{IntParam, Param, ParamMap,
ParamValidators}
+import org.apache.spark.ml.param.shared.{HasLabelCol, HasPredictionCol}
+import org.apache.spark.ml.util.{DefaultParamsReadable,
DefaultParamsWritable, Identifiable}
+import org.apache.spark.sql.{DataFrame, Dataset, Row, SQLContext}
+import org.apache.spark.sql.functions._
+
+/**
+ * :: Experimental ::
+ * Evaluator for ranking, which expects two input columns: prediction and
label.
+ */
+@Since("2.0.0")
+@Experimental
+final class RankingEvaluator[T: ClassTag] @Since("2.0.0") (@Since("2.0.0")
override val uid: String)
+ extends Evaluator with HasPredictionCol with HasLabelCol with
DefaultParamsWritable with Logging {
+
+ @Since("2.0.0")
+ def this() = this(Identifiable.randomUID("rankingEval"))
+
+ @Since("2.0.0")
+ final val k = new IntParam(this, "k", "Top-K cutoff", (x: Int) => x > 0)
+
+ /** @group getParam */
+ @Since("2.0.0")
+ def getK: Int = $(k)
+
+ /** @group setParam */
+ @Since("2.0.0")
+ def setK(value: Int): this.type = set(k, value)
+
+ setDefault(k -> 1)
+
+ /**
+ * Param for metric name in evaluation. Supports:
+ * - `"map"` (default): Mean Average Precision
+ * - `"mapk"`: Mean Average Precision@K
+ * - `"ndcg"`: Normalized Discounted Cumulative Gain
+ * - `"mrr"`: Mean Reciprocal Rank
+ *
+ * @group param
+ */
+ @Since("2.0.0")
+ val metricName: Param[String] = {
+ val allowedParams = ParamValidators.inArray(Array("map", "mapk",
"ndcg", "mrr"))
+ new Param(this, "metricName", "metric name in evaluation
(map|mapk|ndcg||mrr)", allowedParams)
+ }
+
+ /** @group getParam */
+ @Since("2.0.0")
+ def getMetricName: String = $(metricName)
+
+ /** @group setParam */
+ @Since("2.0.0")
+ def setMetricName(value: String): this.type = set(metricName, value)
+
+ /** @group setParam */
+ @Since("2.0.0")
+ def setPredictionCol(value: String): this.type = set(predictionCol,
value)
+
+ /** @group setParam */
+ @Since("2.0.0")
+ def setLabelCol(value: String): this.type = set(labelCol, value)
+
+ setDefault(metricName -> "map")
+
+ @Since("2.0.0")
+ override def evaluate(dataset: Dataset[_]): Double = {
+ val schema = dataset.schema
+ val predictionColName = $(predictionCol)
+ val predictionType = schema($(predictionCol)).dataType
+ val labelColName = $(labelCol)
+ val labelType = schema($(labelCol)).dataType
+ require(predictionType == labelType,
+ s"Prediction column $predictionColName and Label column
$labelColName " +
+ s"must be of the same type, but Prediction column
$predictionColName is $predictionType " +
+ s"and Label column $labelColName is $labelType")
+
+ val metric = $(metricName) match {
+ case "map" => meanAveragePrecision(dataset)
+ case "ndcg" => normalizedDiscountedCumulativeGain(dataset)
+ case "mapk" => meanAveragePrecisionAtK(dataset)
+ case "mrr" => meanReciprocalRank(dataset)
+ }
+ metric
+ }
+
+ /**
+ * Returns the mean average precision (MAP) of all the queries.
+ * If a query has an empty ground truth set, the average precision will
be zero and a log
+ * warning is generated.
+ */
+ private def meanAveragePrecision(dataset: Dataset[_]): Double = {
+ val sc = SparkContext.getOrCreate()
+ val sqlContext = SQLContext.getOrCreate(sc)
+ import sqlContext.implicits._
+
+ dataset.map{ case (prediction: Array[T], label: Array[T]) =>
+ val labSet = label.toSet
+
+ if (labSet.nonEmpty) {
+ var i = 0
+ var cnt = 0
+ var precSum = 0.0
+ val n = prediction.length
+ while (i < n) {
+ if (labSet.contains(prediction(i))) {
+ cnt += 1
+ precSum += cnt.toDouble / (i + 1)
+ }
+ i += 1
+ }
+ precSum / labSet.size
+ } else {
+ logWarning("Empty ground truth set, check input data")
+ 0.0
+ }
+ }.reduce{ (a, b) => a + b } / dataset.count
+ }
+
+ /**
+ * Compute the average NDCG value of all the queries, truncated at
ranking position k.
+ * The discounted cumulative gain at position k is computed as:
+ * sum,,i=1,,^k^ (2^{relevance of ''i''th item}^ - 1) / log(i + 1),
+ * and the NDCG is obtained by dividing the DCG value on the ground
truth set. In the current
+ * implementation, the relevance value is binary.
+
+ * If a query has an empty ground truth set, zero will be used as ndcg
together with
+ * a log warning.
+ *
+ * See the following paper for detail:
+ *
+ * IR evaluation methods for retrieving highly relevant documents. K.
Jarvelin and J. Kekalainen
+ */
+ private def normalizedDiscountedCumulativeGain(dataset: Dataset[_]):
Double = {
+ val sc = SparkContext.getOrCreate()
+ val sqlContext = SQLContext.getOrCreate(sc)
+ import sqlContext.implicits._
+
+ dataset.map{ case (prediction: Array[T], label: Array[T]) =>
+ val labSet = label.toSet
+
+ if (labSet.nonEmpty) {
+ val labSetSize = labSet.size
+ val n = math.min(math.max(prediction.length, labSetSize), $(k))
+ var maxDcg = 0.0
+ var dcg = 0.0
+ var i = 0
+ while (i < n) {
+ val gain = 1.0 / math.log(i + 2)
+ if (labSet.contains(prediction(i))) {
+ dcg += gain
+ }
+ if (i < labSetSize) {
+ maxDcg += gain
--- End diff --
Yeah the counter-argument is that, in the binary case, the best result set
is almost surely all relevant, so this is only normalizing away the discount
factor to put this on a scale of 0 to 1, and maybe make results comparable
across different k. I thought that was its purpose, really.
The latter interpretation is strictly evaluating ranking within the
returned results, and so doesn't measure whether the algorithm returned good
results to begin with (doesn't measure anything about how the returned results
are ranked relative to everything else). Failing to return a relevant result
doesn't penalize you, which seems wrong.
The former interpretation is measuring something about both, which is why
that version had always made sense to me as a metric
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
