net.sourceforge.openforecast.models

Class WeightedMovingAverageModel

java.lang.Object

net.sourceforge.openforecast.models.AbstractForecastingModel

net.sourceforge.openforecast.models.AbstractTimeBasedModel

net.sourceforge.openforecast.models.WeightedMovingAverageModel

All Implemented Interfaces:

ForecastingModel

Direct Known Subclasses:

MovingAverageModel

public class WeightedMovingAverageModel
extends AbstractTimeBasedModel

A weighted moving average forecast model is based on an artificially constructed time series in which the value for a given time period is replaced by the weighted mean of that value and the values for some number of preceding time periods. As you may have guessed from the description, this model is best suited to time-series data; i.e. data that changes over time.

Since the forecast value for any given period is a weighted average of the previous periods, then the forecast will always appear to "lag" behind either increases or decreases in the observed (dependent) values. For example, if a data series has a noticable upward trend then a weighted moving average forecast will generally provide an underestimate of the values of the dependent variable.

The weighted moving average model, like the moving average model, has an advantage over other forecasting models in that it does smooth out peaks and troughs (or valleys) in a set of observations. However, like the moving average model, it also has several disadvantages. In particular this model does not produce an actual equation. Therefore, it is not all that useful as a medium-long range forecasting tool. It can only reliably be used to forecast a few periods into the future.

Since:

0.4

Constructor Summary

Constructors

Constructor and Description
WeightedMovingAverageModel(double[] weights) Constructs a new weighted moving average forecasting model, using the specified weights.
WeightedMovingAverageModel(java.lang.String independentVariable, double[] weights) Deprecated. As of 0.4, replaced by WeightedMovingAverageModel(double[]).

Method Summary

Modifier and Type	Method and Description
java.lang.String	getForecastType() Returns a one or two word name of this type of forecasting model.
int	getNumberOfPredictors() Returns the number of predictors used by the underlying model.
java.lang.String	toString() This should be overridden to provide a textual description of the current forecasting model including, where possible, any derived parameters used.

Methods inherited from class net.sourceforge.openforecast.models.AbstractTimeBasedModel

forecast, getIndependentVariable, getMaximumTimeValue, getMinimumTimeValue, getTimeVariable, init

Methods inherited from class net.sourceforge.openforecast.models.AbstractForecastingModel

forecast, getAIC, getBias, getMAD, getMAPE, getMSE, getSAE

Methods inherited from class java.lang.Object

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

Constructor Detail

WeightedMovingAverageModel

public WeightedMovingAverageModel(double[] weights)

Constructs a new weighted moving average forecasting model, using the specified weights. For a valid model to be constructed, you should call init and pass in a data set containing a series of data points with the time variable initialized to identify the independent variable.

The size of the weights array is used to determine the number of observations to be used to calculate the weighted moving average. Additionally, the most recent period will be given the weight defined by the first element of the array; i.e. weights[0].

The size of the weights array is also used to determine the amount of future periods that can effectively be forecast. With a 50 day weighted moving average, then we cannot reasonably - with any degree of accuracy - forecast more than 50 days beyond the last period for which data is available. Even forecasting near the end of this range is likely be be unreliable.

Note on weights

In general, the weights passed to this constructor should add up to 1.0. However, as a convenience, if the sum of the weights does not add up to 1.0, this implementation scales all weights proportionally so that they do sum to 1.0.

Parameters:

weights - an array of weights to assign to the historical observations when calculating the weighted moving average.

WeightedMovingAverageModel

public WeightedMovingAverageModel(java.lang.String independentVariable,
                                  double[] weights)

Deprecated. As of 0.4, replaced by WeightedMovingAverageModel(double[]).

Constructs a new weighted moving average forecasting model, using the named variable as the independent variable and the specified weights.

Parameters:

independentVariable - the name of the independent variable to use in this model.

weights - an array of weights to assign to the historical observations when calculating the weighted moving average.

Method Detail

getNumberOfPredictors

public int getNumberOfPredictors()

Returns the number of predictors used by the underlying model.

Returns:

the number of predictors used by the underlying model.

getForecastType

public java.lang.String getForecastType()

Returns a one or two word name of this type of forecasting model. Keep this short. A longer description should be implemented in the toString method.

Specified by:

getForecastType in interface ForecastingModel

Overrides:

getForecastType in class AbstractTimeBasedModel

Returns:

a string representation of the type of forecasting model implemented.

toString

public java.lang.String toString()

This should be overridden to provide a textual description of the current forecasting model including, where possible, any derived parameters used.

Specified by:

toString in interface ForecastingModel

Overrides:

toString in class AbstractTimeBasedModel

Returns:

a string representation of the current forecast model, and its parameters.