org.openimaj.ml.timeseries.aggregator

Class WindowedLinearRegressionAggregator

java.lang.Object

org.openimaj.ml.timeseries.aggregator.WindowedLinearRegressionAggregator

All Implemented Interfaces:

SynchronisedTimeSeriesCollectionAggregator<DoubleTimeSeries,DoubleSynchronisedTimeSeriesCollection,DoubleTimeSeries>, TimeSeriesCollectionAggregator<DoubleTimeSeries,DoubleSynchronisedTimeSeriesCollection,DoubleTimeSeries>

public class WindowedLinearRegressionAggregator
extends Object
implements SynchronisedTimeSeriesCollectionAggregator<DoubleTimeSeries,DoubleSynchronisedTimeSeriesCollection,DoubleTimeSeries>

An implementation of a general linear regressive such that the values of a timeseries Y are predicted using the values of a set of time series X at some offset over some time window. X may potentially contain Y itself which turns this into an auto-regressive model augmented with extra information. Furthermore, varying window sizes and offsets may be used for each time series X. This is all achieved with SynchronisedTimeSeriesCollection which model a set of timeseries which are synchronised. When intitalised, the Y time series must be explicitly specified. By default the

Author:

Sina Samangooei (ss@ecs.soton.ac.uk)

Constructor Summary

Constructors

Constructor and Description
WindowedLinearRegressionAggregator(String ydataName) Calculate the regression from the same time series inputed
WindowedLinearRegressionAggregator(String ydataName, boolean autoregressive) Calculate the regression from the same time series inputed
WindowedLinearRegressionAggregator(String ydataName, boolean autoregressive, IndependentPair<Integer,Integer>... windowOffsets) Perform regression s.t.
WindowedLinearRegressionAggregator(String ydataName, int windowsize) Perform regression s.t.
WindowedLinearRegressionAggregator(String ydataName, int windowsize, boolean autoregressive) Perform regression s.t.
WindowedLinearRegressionAggregator(String ydataName, int windowsize, int offset) Perform regression s.t.
WindowedLinearRegressionAggregator(String ydataName, int windowsize, int offset, boolean autoregressive) Perform regression s.t.
WindowedLinearRegressionAggregator(String ydataName, int windowsize, int offset, boolean autoregressive, DoubleSynchronisedTimeSeriesCollection other) Perform regression s.t.

Method Summary

Modifier and Type	Method and Description
DoubleTimeSeries	aggregate(DoubleSynchronisedTimeSeriesCollection series)
LinearRegression	getReg()
LinearRegression	getRegression()

Methods inherited from class java.lang.Object

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

Constructor Detail

WindowedLinearRegressionAggregator

public WindowedLinearRegressionAggregator(String ydataName)

Calculate the regression from the same time series inputed

Parameters:

ydataName -

WindowedLinearRegressionAggregator

public WindowedLinearRegressionAggregator(String ydataName,
                                          boolean autoregressive)

Calculate the regression from the same time series inputed

Parameters:

ydataName -

autoregressive - whether the ydata should be used in regression

WindowedLinearRegressionAggregator

public WindowedLinearRegressionAggregator(String ydataName,
                                          int windowsize)

Perform regression s.t. Y = Sum(w_{0-i} * x_{0-i}) + c using the same window size for all other time series

Parameters:

ydataName -

windowsize -

WindowedLinearRegressionAggregator

public WindowedLinearRegressionAggregator(String ydataName,
                                          int windowsize,
                                          boolean autoregressive)

Perform regression s.t. Y = Sum(w_{0-i} * x_{0-i}) + c using the same window size for all other time series

Parameters:

ydataName -

windowsize -

autoregressive - whether the ydata should be used in regression

WindowedLinearRegressionAggregator

public WindowedLinearRegressionAggregator(String ydataName,
                                          int windowsize,
                                          int offset)

Perform regression s.t. y = Sum(w_{0-i} * x_{0-i}) + c for i from 1 to windowsize with some offset. The same windowsize and offset is used for each time series

Parameters:

ydataName -

windowsize -

offset -

WindowedLinearRegressionAggregator

public WindowedLinearRegressionAggregator(String ydataName,
                                          int windowsize,
                                          int offset,
                                          boolean autoregressive)

Perform regression s.t. y = Sum(w_{0-i} * x_{0-i}) + c for i from 1 to windowsize with some offset. The same windowsize and offset is used for each time series

Parameters:

ydataName -

windowsize -

offset -

autoregressive - whether the ydata should be used in regression

WindowedLinearRegressionAggregator

public WindowedLinearRegressionAggregator(String ydataName,
                                          int windowsize,
                                          int offset,
                                          boolean autoregressive,
                                          DoubleSynchronisedTimeSeriesCollection other)

Perform regression s.t. y = Sum(w_{0-i} * x_{0-i}) + c for i from 1 to windowsize with some offset. The same windowsize and offset is used for each time series

Parameters:

ydataName -

windowsize -

offset -

autoregressive - whether the ydata should be used in regression

other -

WindowedLinearRegressionAggregator

@SafeVarargs
public WindowedLinearRegressionAggregator(String ydataName,
                                                       boolean autoregressive,
                                                       IndependentPair<Integer,Integer>... windowOffsets)

Perform regression s.t. y = Sum(w_{0-i} * x_{0-i}) + c for i from 1 to windowsize with some offset. The same windowsize and offset is used for each time series

Parameters:

ydataName -

autoregressive - whether the ydata should be used in regression

windowOffsets -

Method Detail

getRegression

public LinearRegression getRegression()

Returns:

the WindowedLinearRegressionAggregator's underlying LinearRegression model

aggregate

public DoubleTimeSeries aggregate(DoubleSynchronisedTimeSeriesCollection series)

Specified by:

aggregate in interface SynchronisedTimeSeriesCollectionAggregator<DoubleTimeSeries,DoubleSynchronisedTimeSeriesCollection,DoubleTimeSeries>

Specified by:

aggregate in interface TimeSeriesCollectionAggregator<DoubleTimeSeries,DoubleSynchronisedTimeSeriesCollection,DoubleTimeSeries>

Parameters:

series - aggregate this collection into the output

Returns:

an aggregation of the input

getReg

public LinearRegression getReg()