I'm trying to create model and evaluate in Apache Spark 3.1.1 with OneR algorithm. I have .csv file with normalized data (all values are double, but some values are very close to 0).

I was reading MLlib main guide OnevsRest and code is very similar to this:

SparkSession session = SparkSession
                .builder()
                .appName("Spark test")
                .master("local")
                .getOrCreate();

JavaRDD<LabeledPoint> data = loadData(session, "path.csv");

LogisticRegression logisticRegression = new LogisticRegression().setMaxIter(20);
OneVsRest oneR = new OneVsRest().setClassifier(logisticRegression);

BinaryClassificationEvaluator evaluator = new BinaryClassificationEvaluator();
MulticlassMetrics metrics = new MulticlassMetrics(data.rdd());
MulticlassClassificationEvaluator multiEvaluator = new MulticlassClassificationEvaluator()
        .setMetricName("accuracy");
        
JavaRDD<LabeledPoint>[] javaRDDS = data.randomSplit(new double[]{0.7, 0.3});
JavaRDD<LabeledPoint> trainingRDD = javaRDDS[0], testRDD = javaRDDS[1];
Dataset<Row> trainingDataset = session.createDataFrame(trainingRDD, LabeledPoint.class);
Dataset<Row> testDataset = session.createDataFrame(testRDD, LabeledPoint.class);

OneVsRestModel oneRModel = oneR.fit(trainingDataset);
Dataset<Row> oneRPredictions = oneRModel.transform(testDataset).select("prediction", "label");

double oneRAcc = evaluator.evaluate(oneRPredictions);
System.out.println("OneR: \r\n");
System.out.println("Accuracy: " + oneRAcc);
System.out.println("--------------------------------");

session.close();

This code throws exception:

Exception in thread "main" java.lang.IllegalArgumentException: requirement failed: rawPredictionCol vectors must have length=2, but got 3
    at scala.Predef$.require(Predef.scala:281)
    at org.apache.spark.ml.evaluation.BinaryClassificationEvaluator.$anonfun$getMetrics$1(BinaryClassificationEvaluator.scala:126)
    at scala.runtime.java8.JFunction1$mcVI$sp.apply(JFunction1$mcVI$sp.java:23)
    at scala.Option.foreach(Option.scala:407)
    at org.apache.spark.ml.evaluation.BinaryClassificationEvaluator.getMetrics(BinaryClassificationEvaluator.scala:126)
    at org.apache.spark.ml.evaluation.BinaryClassificationEvaluator.evaluate(BinaryClassificationEvaluator.scala:100)
    at Classification.main(Classification.java:64)

Why this code doesn't work? I thought that problem is in .select("prediction", "label") because I don't know what I have in Dataset<Row> after transform, but vectors must have length of 2 not 3 is strange. I'm trying multiclass classification with 3 classes.

Edit

I was using BinaryClassificationEvaluator evaluator instead of MulticlassClassificationEvaluator multiEvaluator by mistake. Now error message makes sense.