Package org.pentaho.di.forecasting

Class WekaForecasting

  • All Implemented Interfaces:
    org.pentaho.di.core.ExtensionDataInterface, org.pentaho.di.core.logging.HasLogChannelInterface, org.pentaho.di.core.logging.LoggingObjectInterface, org.pentaho.di.core.logging.LoggingObjectLifecycleInterface, org.pentaho.di.core.variables.VariableSpace, org.pentaho.di.trans.step.StepInterface
    public class WekaForecasting
extends org.pentaho.di.trans.step.BaseStep
implements org.pentaho.di.trans.step.StepInterface
    Applies a pre-built weka forecasting model to make a forecast for future time steps. Incoming rows are treated as historical data for priming the model with - i.e. the priming process ensures that all the lagged and derived field values for the forecaster are filled with respect to the most recent historical data. The forecaster can then be applied in a closed-loop manner to forecast for future time steps. The closed-loop process takes values forecasted for the next time step and feeds them back into the model in order to make a forecast for subsequent time step, and so on. Therefore, the priming data is expected to up to, and including, the current time step (i.e. up to one step immediately prior to the first future forecast step). IMPORTANT: priming rows are assumed to be equally spaced in time and are sorted in ascending order of time. If a time stamp is included in the data then we check that this is the case. It is also assumed that the priming data (and overlay data if provided) have the same time interval and periodicity as the data used to train the model. If this is not the case then results may be nonsensical. We do not check for this.

    If the forecasting model has been trained with "overlay" data, i.e. fields that are not forecasted or derived automatically, then the incoming data stream needs to contain these fields - not only for the priming data, but *also* for time steps that are to be forecasted. We assume that overlay data for future time steps to be forecasted is indicated by the presence of rows that contain all missing (null) values for the target fields to be forecasted. Of course, the values of the overlay fields need to be non-missing in this data. The number of "overlay" rows determines the number of steps forecasted by the forecaster, and overides any value specified in the "numSteps" parameter.

    When a forecast is generated, confidence intervals will be included for any steps for which they were estimated when training the model.

    Attributes that the Weka model was constructed from are automatically mapped to incoming Kettle fields on the basis of name and type. Any attributes that cannot be mapped due to type mismatch or not being present in the incoming fields receive missing values when incoming Kettle rows are converted to Weka's Instance format. Similarly, any values for string fields that have not been seen during the training of the Weka model are converted to missing values.

    Version:
    $Revision$
    Author:
    Mark Hall (mhall{[at]}pentaho{[dot]}com)

    • Constructor Summary

      Constructors 
      Constructor Description
      WekaForecasting​(org.pentaho.di.trans.step.StepMeta stepMeta, org.pentaho.di.trans.step.StepDataInterface stepDataInterface, int copyNr, org.pentaho.di.trans.TransMeta transMeta, org.pentaho.di.trans.Trans trans)
      Creates a new WekaForecasting instance.

    • Method Summary

      All Methods Instance Methods Concrete Methods 
      Modifier and Type Method Description
      boolean init​(org.pentaho.di.trans.step.StepMetaInterface smi, org.pentaho.di.trans.step.StepDataInterface sdi)
      Initialize the step.
      boolean processRow​(org.pentaho.di.trans.step.StepMetaInterface smi, org.pentaho.di.trans.step.StepDataInterface sdi)
      Process an incoming row of data.
      void run()
      Run is where the action happens!

      • Methods inherited from class org.pentaho.di.trans.step.BaseStep

        addResultFile, addRowListener, addRowSetToInputRowSets, addRowSetToOutputRowSets, addStepListener, batchComplete, beforeStartProcessing, buildLog, canProcessOneRow, checkFeedback, cleanup, clearInputRowSets, clearOutputRowSets, closeQuietly, copyVariablesFrom, decrementLinesRead, decrementLinesWritten, dispatch, dispose, environmentSubstitute, environmentSubstitute, environmentSubstitute, fieldSubstitute, findInputRowSet, findInputRowSet, findOutputRowSet, findOutputRowSet, getBooleanValueOfVariable, getClusterSize, getContainerObjectId, getCopy, getCurrentInputRowSetNr, getCurrentOutputRowSetNr, getDispatcher, getErrorRowMeta, getErrors, getExtensionDataMap, getFilename, getFirstInputRowSet, getInputRowMeta, getInputRowSets, getLinesInput, getLinesOutput, getLinesRead, getLinesRejected, getLinesSkipped, getLinesUpdated, getLinesWritten, getLogChannel, getLogChannelId, getLogFields, getLogLevel, getMetaStore, getNextClassNr, getObjectCopy, getObjectId, getObjectName, getObjectRevision, getObjectType, getOutputRowSets, getParent, getParentVariableSpace, getPartitionID, getPartitionTargets, getPreviewRowMeta, getProcessed, getRegistrationDate, getRemoteInputSteps, getRemoteOutputSteps, getRepartitioning, getRepository, getRepositoryDirectory, getResultFiles, getRow, getRowFrom, getRowHandler, getRowListeners, getRuntime, getServerSockets, getSlaveNr, getSocketRepository, getStatus, getStatusDescription, getStepDataInterface, getStepID, getStepListeners, getStepMeta, getStepMetaInterface, getStepname, getTrans, getTransMeta, getTypeId, getUniqueStepCountAcrossSlaves, getUniqueStepNrAcrossSlaves, getVariable, getVariable, handleGetRowFrom, handlePutRowTo, identifyErrorOutput, incrementLinesInput, incrementLinesOutput, incrementLinesRead, incrementLinesRejected, incrementLinesSkipped, incrementLinesUpdated, incrementLinesWritten, initBeforeStart, initializeVariablesFrom, injectVariables, isBasic, isDebug, isDetailed, isDistributed, isForcingSeparateLogging, isGatheringMetrics, isInitialising, isLoggingObjectInUse, isMapping, isPartitioned, isPaused, isRowLevel, isRunning, isSafeStopped, isStopped, isUsingThreadPriorityManagment, listVariables, logBasic, logBasic, logDebug, logDebug, logDetailed, logDetailed, logError, logError, logError, logMinimal, logMinimal, logRowlevel, logRowlevel, logSummary, markStart, markStop, openRemoteInputStepSocketsOnce, openRemoteOutputStepSocketsOnce, outputIsDone, pauseRunning, putError, putRow, putRowTo, removeRowListener, resumeRunning, rowsetInputSize, rowsetOutputSize, safeModeChecking, safeModeChecking, setCarteObjectId, setCopy, setCurrentInputRowSetNr, setCurrentOutputRowSetNr, setDistributed, setErrorRowMeta, setErrors, setForcingSeparateLogging, setGatheringMetrics, setInputRowMeta, setInputRowSets, setInternalVariables, setLinesInput, setLinesOutput, setLinesRead, setLinesRejected, setLinesSkipped, setLinesUpdated, setLinesWritten, setLoggingObjectInUse, setLogLevel, setMetaStore, setOutputDone, setOutputRowSets, setParentVariableSpace, setPartitioned, setPartitionID, setPartitionTargets, setPaused, setPaused, setPreviewRowMeta, setRepartitioning, setRepository, setRowHandler, setRunning, setSafeStopped, setServerSockets, setSocketRepository, setStepDataInterface, setStepListeners, setStepMeta, setStepMetaInterface, setStepname, setStopped, setTransMeta, setUsingThreadPriorityManagment, setVariable, shareVariablesWith, stopAll, stopRunning, stopRunning, swapFirstInputRowSetIfExists, toString, verifyInputDeadLock, waitUntilTransformationIsStarted

      • Methods inherited from interface org.pentaho.di.core.logging.LoggingObjectLifecycleInterface

        callAfterLog, callBeforeLog

      • Methods inherited from interface org.pentaho.di.trans.step.StepInterface

        addRowListener, addRowSetToInputRowSets, addRowSetToOutputRowSets, addStepListener, afterFinishProcessing, batchComplete, beforeStartProcessing, canProcessOneRow, cleanup, dispose, getCopy, getCurrentInputRowSetNr, getCurrentOutputRowSetNr, getErrors, getInputRowSets, getLinesInput, getLinesOutput, getLinesRead, getLinesRejected, getLinesUpdated, getLinesWritten, getLogChannel, getMetaStore, getOutputRowSets, getPartitionID, getProcessed, getRepository, getResultFiles, getRow, getRowListeners, getRuntime, getStatus, getStepID, getStepMeta, getStepname, getTrans, identifyErrorOutput, initBeforeStart, isMapping, isPartitioned, isPaused, isRunning, isSafeStopped, isStopped, isUsingThreadPriorityManagment, markStart, markStop, pauseRunning, putRow, removeRowListener, resumeRunning, rowsetInputSize, rowsetOutputSize, setCurrentInputRowSetNr, setCurrentOutputRowSetNr, setErrors, setLinesRejected, setMetaStore, setOutputDone, setPartitioned, setPartitionID, setRepartitioning, setRepository, setRunning, setSafeStopped, setStopped, setUsingThreadPriorityManagment, stopAll, stopRunning, subStatuses

      • Methods inherited from interface org.pentaho.di.core.variables.VariableSpace

        copyVariablesFrom, environmentSubstitute, environmentSubstitute, environmentSubstitute, fieldSubstitute, getBooleanValueOfVariable, getParentVariableSpace, getVariable, getVariable, initializeVariablesFrom, injectVariables, listVariables, setParentVariableSpace, setVariable, shareVariablesWith

    • Field Detail

      • m_overlayData

        protected List<Object[]> m_overlayData

      • m_bufferedPrimeData

        protected List<Object[]> m_bufferedPrimeData

      • m_isIncrementallyPrimeable

        protected boolean m_isIncrementallyPrimeable

      • m_isUsingOverlayData

        protected boolean m_isUsingOverlayData

      • m_rebuildModel

        protected boolean m_rebuildModel
        rebuild the model on the incoming data before forecasting?

      • m_modelLagMaker

        protected weka.classifiers.timeseries.core.TSLagMaker m_modelLagMaker

      • m_timeStampName

        protected String m_timeStampName

      • m_timeStampRowIndex

        protected int m_timeStampRowIndex

      • m_fieldsToForecast

        protected List<String> m_fieldsToForecast

    • Constructor Detail

      • WekaForecasting

        public WekaForecasting​(org.pentaho.di.trans.step.StepMeta stepMeta,
                       org.pentaho.di.trans.step.StepDataInterface stepDataInterface,
                       int copyNr,
                       org.pentaho.di.trans.TransMeta transMeta,
                       org.pentaho.di.trans.Trans trans)
        Creates a new WekaForecasting instance.
        Parameters:
        stepMeta - holds the step's meta data
        stepDataInterface - holds the step's temporary data
        copyNr - the number assigned to the step
        transMeta - meta data for the transformation
        trans - a Trans value

    • Method Detail

      • processRow

        public boolean processRow​(org.pentaho.di.trans.step.StepMetaInterface smi,
                          org.pentaho.di.trans.step.StepDataInterface sdi)
                   throws org.pentaho.di.core.exception.KettleException
        Process an incoming row of data.
        Specified by:
        processRow in interface org.pentaho.di.trans.step.StepInterface
        Overrides:
        processRow in class org.pentaho.di.trans.step.BaseStep
        Parameters:
        smi - a StepMetaInterface value
        sdi - a StepDataInterface value
        Returns:
        a boolean value
        Throws:
        org.pentaho.di.core.exception.KettleException - if an error occurs

      • init

        public boolean init​(org.pentaho.di.trans.step.StepMetaInterface smi,
                    org.pentaho.di.trans.step.StepDataInterface sdi)
        Initialize the step.
        Specified by:
        init in interface org.pentaho.di.trans.step.StepInterface
        Overrides:
        init in class org.pentaho.di.trans.step.BaseStep
        Parameters:
        smi - a StepMetaInterface value
        sdi - a StepDataInterface value
        Returns:
        a boolean value

      • run

        public void run()
        Run is where the action happens!