Package org.elasticsearch.search.aggregations.pipeline

Class MovingFunctions

java.lang.Object

org.elasticsearch.search.aggregations.pipeline.MovingFunctions

public class MovingFunctions
extends java.lang.Object

Provides a collection of static utility methods that can be referenced from MovingFunction script contexts

Constructor Summary

Constructors

Constructor	Description
MovingFunctions()

Method Summary

Modifier and Type	Method	Description
static double	ewma(double[] values, double alpha)	Calculate a exponentially weighted moving average.
static double	holt(double[] values, double alpha, double beta)	Calculate a doubly exponential weighted moving average Alpha controls the smoothing of the data.
static double[]	holtForecast(double[] values, double alpha, double beta, int numForecasts)	Version of holt that can "forecast", not exposed as a whitelisted function for moving_fn scripts, but here as compatibility/code sharing for existing moving_avg agg.
static double	holtWinters(double[] values, double alpha, double beta, double gamma, int period, boolean multiplicative)	Calculate a triple exponential weighted moving average Alpha controls the smoothing of the data.
static double[]	holtWintersForecast(double[] values, double alpha, double beta, double gamma, int period, double padding, boolean multiplicative, int numForecasts)	Version of holt-winters that can "forecast", not exposed as a whitelisted function for moving_fn scripts, but here as compatibility/code sharing for existing moving_avg agg.
static double	linearWeightedAvg(double[] values)	Calculate a linearly weighted moving average, such that older values are linearly less important.
static double	max(double[] values)	Find the maximum value in a window of values.
static double	min(double[] values)	Find the minimum value in a window of values If all values are missing/null/NaN, the return value will be NaN
static double	stdDev(double[] values, double avg)	Calculate a standard deviation over the values using the provided average.
static double	sum(double[] values)	Find the sum of a window of values If all values are missing/null/NaN, the return value will be 0.0
static double	unweightedAvg(double[] values)	Calculate a simple unweighted (arithmetic) moving average.

Methods inherited from class java.lang.Object

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

Constructor Detail

MovingFunctions

public MovingFunctions()

Method Detail

max

public static double max(double[] values)

Find the maximum value in a window of values. If all values are missing/null/NaN, the return value will be NaN

min

public static double min(double[] values)

Find the minimum value in a window of values If all values are missing/null/NaN, the return value will be NaN

sum

public static double sum(double[] values)

Find the sum of a window of values If all values are missing/null/NaN, the return value will be 0.0

unweightedAvg

public static double unweightedAvg(double[] values)

Calculate a simple unweighted (arithmetic) moving average. Only finite values are averaged. NaN or null are ignored. If all values are missing/null/NaN, the return value will be NaN. The average is based on the count of non-null, non-NaN values.

stdDev

public static double stdDev(double[] values,
                            double avg)

Calculate a standard deviation over the values using the provided average. Only finite values are averaged. NaN or null are ignored. If all values are missing/null/NaN, the return value will be NaN. The average is based on the count of non-null, non-NaN values.

linearWeightedAvg

public static double linearWeightedAvg(double[] values)

Calculate a linearly weighted moving average, such that older values are linearly less important. "Time" is determined by position in collection Only finite values are averaged. NaN or null are ignored. If all values are missing/null/NaN, the return value will be NaN The average is based on the count of non-null, non-NaN values.

ewma

public static double ewma(double[] values,
                          double alpha)

Calculate a exponentially weighted moving average. Alpha controls the smoothing of the data. Alpha = 1 retains no memory of past values (e.g. a random walk), while alpha = 0 retains infinite memory of past values (e.g. the series mean). Useful values are somewhere in between. Defaults to 0.5. Only finite values are averaged. NaN or null are ignored. If all values are missing/null/NaN, the return value will be NaN The average is based on the count of non-null, non-NaN values.

Parameters:

alpha - A double between 0-1 inclusive, controls data smoothing

holt

public static double holt(double[] values,
                          double alpha,
                          double beta)

Calculate a doubly exponential weighted moving average Alpha controls the smoothing of the data. Alpha = 1 retains no memory of past values (e.g. a random walk), while alpha = 0 retains infinite memory of past values (e.g. the series mean). Useful values are somewhere in between. Defaults to 0.5. Beta is equivalent to alpha, but controls the smoothing of the trend instead of the data Only finite values are averaged. NaN or null are ignored. If all values are missing/null/NaN, the return value will be NaN The average is based on the count of non-null, non-NaN values.

Parameters:

alpha - A double between 0-1 inclusive, controls data smoothing

beta - a double between 0-1 inclusive, controls trend smoothing

holtForecast

public static double[] holtForecast(double[] values,
                                    double alpha,
                                    double beta,
                                    int numForecasts)

Version of holt that can "forecast", not exposed as a whitelisted function for moving_fn scripts, but here as compatibility/code sharing for existing moving_avg agg. Can be removed when moving_avg is gone.

holtWinters

public static double holtWinters(double[] values,
                                 double alpha,
                                 double beta,
                                 double gamma,
                                 int period,
                                 boolean multiplicative)

Calculate a triple exponential weighted moving average Alpha controls the smoothing of the data. Alpha = 1 retains no memory of past values (e.g. a random walk), while alpha = 0 retains infinite memory of past values (e.g. the series mean). Useful values are somewhere in between. Defaults to 0.5. Beta is equivalent to alpha, but controls the smoothing of the trend instead of the data. Gamma is equivalent to alpha, but controls the smoothing of the seasonality instead of the data Only finite values are averaged. NaN or null are ignored. If all values are missing/null/NaN, the return value will be NaN The average is based on the count of non-null, non-NaN values.

Parameters:

alpha - A double between 0-1 inclusive, controls data smoothing

beta - a double between 0-1 inclusive, controls trend smoothing

gamma - a double between 0-1 inclusive, controls seasonality smoothing

period - the expected periodicity of the data

multiplicative - true if multiplicative HW should be used. False for additive

holtWintersForecast

public static double[] holtWintersForecast(double[] values,
                                           double alpha,
                                           double beta,
                                           double gamma,
                                           int period,
                                           double padding,
                                           boolean multiplicative,
                                           int numForecasts)

Version of holt-winters that can "forecast", not exposed as a whitelisted function for moving_fn scripts, but here as compatibility/code sharing for existing moving_avg agg. Can be removed when moving_avg is gone.