kieker.tools.tslib.forecast.arima
Class ARIMAForecaster

java.lang.Object
  kieker.tools.tslib.forecast.AbstractForecaster<Double>
      kieker.tools.tslib.forecast.AbstractRForecaster
          kieker.tools.tslib.forecast.arima.ARIMAForecaster

All Implemented Interfaces:

IForecaster<Double>

public class ARIMAForecaster
extends AbstractRForecaster

An R-based time series forecaster, auto-arima model selection. This is one of the forecasters used in the research paper on Self-adaptive workload classification and forecasting for proactive resource provisioning (http://dx.doi.org/10.1002/cpe.3224), authored by Herbst et al.

Since:

1.10

Author:

Andre van Hoorn, Nikolas Herbst The automated ARIMA model selection process of the R forecasting package starts with a complex estimation of an appropriate ARIMA(p, d, q)(P, D, Q)m model by using unit-root tests and an information criterions (like the AIC) in combination with a step-wise procedure for traversing a relevant model space. The selected ARIMA model is then fitted to the data to provide point forecasts and confidence intervals.

Constructor Summary

ARIMAForecaster(ITimeSeries<Double> historyTimeseries)

ARIMAForecaster(ITimeSeries<Double> historyTimeseries, int confidenceLevel)

Method Summary

Methods inherited from class kieker.tools.tslib.forecast.AbstractRForecaster

forecast, removeNullValues

Methods inherited from class kieker.tools.tslib.forecast.AbstractForecaster

getConfidenceLevel, getTsOriginal

Methods inherited from class java.lang.Object

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

Constructor Detail

ARIMAForecaster

public ARIMAForecaster(ITimeSeries<Double> historyTimeseries)

Parameters:

historyTimeseries - timeseries used by forecating algo

ARIMAForecaster

public ARIMAForecaster(ITimeSeries<Double> historyTimeseries,
                       int confidenceLevel)

Parameters:

historyTimeseries - timeseries used by forecating algo

confidenceLevel - value of confidence