本文主要对 Spark ML库下模型评估指标的讲解，以下代码均以Jupyter Notebook进行讲解，Spark版本为2.4.5。模型评估指标位于包org.apache.spark.ml.evaluation下。
模型评估指标是指测试集的评估指标，而不是训练集的评估指标
1、回归评估指标
RegressionEvaluator
Evaluator for regression, which expects two input columns: prediction and label.
评估指标支持以下几种：
val metricName: Param[String]
"rmse"(default): root mean squared error"mse": mean squared error"r2": R2 metric"mae": mean absolute error
Examples
# import dependencies
import org.apache.spark.ml.regression.LinearRegression
import org.apache.spark.ml.evaluation.RegressionEvaluator
// Load training data
val data = spark.read.format("libsvm")
.load("/data1/software/spark/data/mllib/sample_linear_regression_data.txt")
val lr = new LinearRegression()
.setMaxIter(10)
.setRegParam(0.3)
.setElasticNetParam(0.8)
// Fit the model
val lrModel = lr.fit(training)
// Summarize the model over the training set and print out some metrics
val trainingSummary = lrModel.summary
println(s"Train MSE: ${trainingSummary.meanSquaredError}")
println(s"Train RMSE: ${trainingSummary.rootMeanSquaredError}")
println(s"Train MAE: ${trainingSummary.meanAbsoluteError}")
println(s"Train r2: ${trainingSummary.r2}")
val predictions = lrModel.transform(test)
// 计算精度
val evaluator = new RegressionEvaluator()
.setLabelCol("label")
.setPredictionCol("prediction")
.setMetricName("mse")
val accuracy = evaluator.evaluate(predictions)
print(s"Test MSE: ${accuracy}")
输出：
Train MSE: 101.57870147367461
Train RMSE: 10.078625971513905
Train MAE: 8.108865602095849
Train r2: 0.039467152584195975
Test MSE: 114.28454406581636
2、分类评估指标
2.1 BinaryClassificationEvaluator
Evaluator for binary classification, which expects two input columns: rawPrediction and label. The rawPrediction column can be of type double (binary 0/1 prediction, or probability of label 1) or of type vector (length-2 vector of raw predictions, scores, or label probabilities).
评估指标支持以下几种：
val metricName: Param[String]
param for metric name in evaluation (supports "areaUnderROC" (default), "areaUnderPR")
Examples
import org.apache.spark.ml.classification.LogisticRegression
import org.apache.spark.ml.evaluation.BinaryClassificationEvaluator
import org.apache.spark.ml.evaluation.MulticlassClassificationEvaluator
// Load training data
val data = spark.read.format("libsvm").load("/data1/software/spark/data/mllib/sample_libsvm_data.txt")
val Array(train, test) = data.randomSplit(Array(0.8, 0.2))
val lr = new LogisticRegression()
.setMaxIter(10)
.setRegParam(0.3)
.setElasticNetParam(0.8)
// Fit the model
val lrModel = lr.fit(train)
// Summarize the model over the training set and print out some metrics
val trainSummary = lrModel.summary
println(s"Train accuracy: ${trainSummary.accuracy}")
println(s"Train weightedPrecision: ${trainSummary.weightedPrecision}")
println(s"Train weightedRecall: ${trainSummary.weightedRecall}")
println(s"Train weightedFMeasure: ${trainSummary.weightedFMeasure}")
val predictions = lrModel.transform(test)
predictions.show(5)
// 模型评估
val evaluator = new BinaryClassificationEvaluator()
.setLabelCol("label")
.setRawPredictionCol("rawPrediction")
.setMetricName("areaUnderROC")
val auc = evaluator.evaluate(predictions)
print(s"Test AUC: ${auc}")
val mulEvaluator = new MulticlassClassificationEvaluator()
.setLabelCol("label")
.setPredictionCol("prediction")
.setMetricName("weightedPrecision")
val precision = evaluator.evaluate(predictions)
print(s"Test weightedPrecision: ${precision}")
输出结果：
Train accuracy: 0.9873417721518988
Train weightedPrecision: 0.9876110961486668
Train weightedRecall: 0.9873417721518987
Train weightedFMeasure: 0.9873124561568825
+-----+--------------------+--------------------+--------------------+----------+
|label| features| rawPrediction| probability|prediction|
+-----+--------------------+--------------------+--------------------+----------+
| 0.0|(692,[122,123,148...|[0.29746771419036...|[0.57382336211209...| 0.0|
| 0.0|(692,[125,126,127...|[0.42262389447949...|[0.60411095396791...| 0.0|
| 0.0|(692,[126,127,128...|[0.74220898710237...|[0.67747871191347...| 0.0|
| 0.0|(692,[126,127,128...|[0.77729372618481...|[0.68509655708828...| 0.0|
| 0.0|(692,[127,128,129...|[0.70928896866149...|[0.67024402884354...| 0.0|
+-----+--------------------+--------------------+--------------------+----------+
Test AUC: 1.0
Test weightedPrecision: 1.0
2.2 MulticlassClassificationEvaluator
Evaluator for multiclass classification, which expects two input columns: prediction and label.
注：既然适用于多分类，当然适用于上面的二分类
评估指标支持如下几种：
val metricName: Param[String]
param for metric name in evaluation (supports "f1" (default), "weightedPrecision", "weightedRecall", "accuracy")
Examples
import org.apache.spark.ml.Pipeline
import org.apache.spark.ml.classification.DecisionTreeClassificationModel
import org.apache.spark.ml.classification.DecisionTreeClassifier
import org.apache.spark.ml.evaluation.MulticlassClassificationEvaluator
import org.apache.spark.ml.feature.{IndexToString, StringIndexer, VectorIndexer}
// Load the data stored in LIBSVM format as a DataFrame.
val data = spark.read.format("libsvm").load("/data1/software/spark/data/mllib/sample_libsvm_data.txt")
// Index labels, adding metadata to the label column.
// Fit on whole dataset to include all labels in index.
val labelIndexer = new StringIndexer()
.setInputCol("label")
.setOutputCol("indexedLabel")
.fit(data)
// Automatically identify categorical features, and index them.
val featureIndexer = new VectorIndexer()
.setInputCol("features")
.setOutputCol("indexedFeatures")
.setMaxCategories(4) // features with > 4 distinct values are treated as continuous.
.fit(data)
// Split the data into training and test sets (30% held out for testing).
val Array(trainingData, testData) = data.randomSplit(Array(0.7, 0.3))
// Train a DecisionTree model.
val dt = new DecisionTreeClassifier()
.setLabelCol("indexedLabel")
.setFeaturesCol("indexedFeatures")
// Convert indexed labels back to original labels.
val labelConverter = new IndexToString()
.setInputCol("prediction")
.setOutputCol("predictedLabel")
.setLabels(labelIndexer.labels)
// Chain indexers and tree in a Pipeline.
val pipeline = new Pipeline()
.setStages(Array(labelIndexer, featureIndexer, dt, labelConverter))
// Train model. This also runs the indexers.
val model = pipeline.fit(trainingData)
// Make predictions.
val predictions = model.transform(testData)
// Select example rows to display.
predictions.select("predictedLabel", "label", "features").show(5)
// Select (prediction, true label) and compute test error.
val evaluator = new MulticlassClassificationEvaluator()
.setLabelCol("indexedLabel")
.setPredictionCol("prediction")
.setMetricName("accuracy")
val accuracy = evaluator.evaluate(predictions)
println(s"Test Error = ${(1.0 - accuracy)}")
输出结果：
+--------------+-----+--------------------+
|predictedLabel|label| features|
+--------------+-----+--------------------+
| 0.0| 0.0|(692,[95,96,97,12...|
| 0.0| 0.0|(692,[122,123,124...|
| 0.0| 0.0|(692,[122,123,148...|
| 0.0| 0.0|(692,[126,127,128...|
| 0.0| 0.0|(692,[126,127,128...|
+--------------+-----+--------------------+
only showing top 5 rows
Test Error = 0.040000000000000036