net.sourceforge.openforecast

Class Forecaster

java.lang.Object

net.sourceforge.openforecast.Forecaster

public class Forecaster
extends java.lang.Object

The Forecaster class is a factory class that obtains the best ForecastingModel for the given data set. The interpretation of the "best" forecasting model can be user selected (bias, MAD, MAPE, MSE, SAE or a blend of these), or left up to the Forecaster. If the interpretation is left up to the Forecaster class then it will evaluate a combination of these other measures and comes up with somewhat of a concensus opinion as the to best model. For more details on the different options available, see EvaluationCriteria.

Method Summary

Modifier and Type	Method and Description
static ForecastingModel	getBestForecast(DataSet dataSet) Obtains the best forecasting model for the given DataSet.
static ForecastingModel	getBestForecast(DataSet dataSet, EvaluationCriteria evalMethod) Obtains the best forecasting model for the given DataSet.

Methods inherited from class java.lang.Object

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

Method Detail

getBestForecast

public static ForecastingModel getBestForecast(DataSet dataSet)

Obtains the best forecasting model for the given DataSet. There is some intelligence built into this method to help it determine which forecasting model is best suited to the data. In particular, it will try applying various forecasting models, using different combinations of independent variables and select the one with the least Sum of Absolute Errors (SAE); i.e. the most accurate one based on historical data.

Parameters:

dataSet - a set of observations on which the given model should be based.

Returns:

the best ForecastingModel for the given data set.

getBestForecast

public static ForecastingModel getBestForecast(DataSet dataSet,
                                               EvaluationCriteria evalMethod)

Obtains the best forecasting model for the given DataSet. To determine which model is best the specified EvaluationCriteria is used - this includes options to use bias, mean absolute deviation (MAD), mean absolute percentage error (MAPE), mean squared error (MSE) and more. For a complete list refer to the final static members defined in the EvaluationCriteria class.

Parameters:

dataSet - a set of observations on which the given model should be based.

evalMethod - specifies how to determine the "best" model; using which EvaluationCriteria.

Returns:

the best ForecastingModel for the given data set.

Since:

0.5

See Also:

EvaluationCriteria