Class StreamingTest

Object
org.apache.spark.mllib.stat.test.StreamingTest
All Implemented Interfaces:
Serializable, org.apache.spark.internal.Logging, scala.Serializable

public class StreamingTest extends Object implements org.apache.spark.internal.Logging, scala.Serializable
Performs online 2-sample significance testing for a stream of (Boolean, Double) pairs. The Boolean identifies which sample each observation comes from, and the Double is the numeric value of the observation.

To address novelty affects, the peacePeriod specifies a set number of initial RDD batches of the DStream to be dropped from significance testing.

The windowSize sets the number of batches each significance test is to be performed over. The window is sliding with a stride length of 1 batch. Setting windowSize to 0 will perform cumulative processing, using all batches seen so far.

Different tests may be used for assessing statistical significance depending on assumptions satisfied by data. For more details, see StreamingTestMethod. The testMethod specifies which test will be used.

Use a builder pattern to construct a streaming test in an application, for example:


   val model = new StreamingTest()
     .setPeacePeriod(10)
     .setWindowSize(0)
     .setTestMethod("welch")
     .registerStream(DStream)
 
See Also:
  • Nested Class Summary

    Nested classes/interfaces inherited from interface org.apache.spark.internal.Logging

    org.apache.spark.internal.Logging.SparkShellLoggingFilter
  • Constructor Summary

    Constructors
    Constructor
    Description
     
  • Method Summary

    Modifier and Type
    Method
    Description
    JavaDStream<org.apache.spark.mllib.stat.test.StreamingTestResult>
    Register a JavaDStream of values for significance testing.
    DStream<org.apache.spark.mllib.stat.test.StreamingTestResult>
    Register a DStream of values for significance testing.
    setPeacePeriod(int peacePeriod)
    Set the number of initial batches to ignore.
    Set the statistical method used for significance testing.
    setWindowSize(int windowSize)
    Set the number of batches to compute significance tests over.

    Methods inherited from class java.lang.Object

    equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

    Methods inherited from interface org.apache.spark.internal.Logging

    initializeForcefully, initializeLogIfNecessary, initializeLogIfNecessary, initializeLogIfNecessary$default$2, isTraceEnabled, log, logDebug, logDebug, logError, logError, logInfo, logInfo, logName, logTrace, logTrace, logWarning, logWarning, org$apache$spark$internal$Logging$$log_, org$apache$spark$internal$Logging$$log__$eq
  • Constructor Details

    • StreamingTest

      public StreamingTest()
  • Method Details

    • registerStream

      public DStream<org.apache.spark.mllib.stat.test.StreamingTestResult> registerStream(DStream<BinarySample> data)
      Register a DStream of values for significance testing.

      Parameters:
      data - stream of BinarySample(key,value) pairs where the key denotes group membership (true = experiment, false = control) and the value is the numerical metric to test for significance
      Returns:
      stream of significance testing results
    • registerStream

      public JavaDStream<org.apache.spark.mllib.stat.test.StreamingTestResult> registerStream(JavaDStream<BinarySample> data)
      Register a JavaDStream of values for significance testing.

      Parameters:
      data - stream of BinarySample(isExperiment,value) pairs where the isExperiment denotes group (true = experiment, false = control) and the value is the numerical metric to test for significance
      Returns:
      stream of significance testing results
    • setPeacePeriod

      public StreamingTest setPeacePeriod(int peacePeriod)
      Set the number of initial batches to ignore. Default: 0.
    • setTestMethod

      public StreamingTest setTestMethod(String method)
      Set the statistical method used for significance testing. Default: "welch"
    • setWindowSize

      public StreamingTest setWindowSize(int windowSize)
      Set the number of batches to compute significance tests over. Default: 0. A value of 0 will use all batches seen so far.
      Parameters:
      windowSize - (undocumented)
      Returns:
      (undocumented)