modl {predtoolsTS}

Building predictive models

Description

This function give us the tools to build predictive models for time series.

Usage

modl(tserie, method = "arima", algorithm = NULL, formula = NULL,
  initialWindow = NULL, horizon = NULL, fixedWindow = NULL)

Arguments

tserie

A ts or prep object.

method

A string. Current methods available are "arima" and "dataMining". Method "arima" is set as default.

algorithm

A string. In case method is "dataMining", pick the algorithm you want to use. There is a complete list of available algorithms here (only regression type allowed): http://topepo.github.io/caret/train-models-by-tag.html.

formula

An integer vector. Contains the indexes from the time series wich will indicate how to extract the features. The last value will be the class index. Default value: c(1:16)

initialWindow

An integer. The initial number of consecutive values in each training set sample. Default value: 30.

horizon

An integer. The number of consecutive values in test set sample. Default value: 15.

fixedWindow

A logical: if FALSE, the training set always start at the first sample and the training set size will vary over data splits. Default value: TRUE.

Details

Returns an object modl which stores all the information related to the final chosen model (errors, parameters, model).

Currently this function covers two different methods: the widely know ARIMA and the "not so used for prediction" data mining. For the data mining we make use of the caret package.

The caret package offers plenty of data mining algorithms. For the data splitting here we use a rolling forecasting origin technique, wich works better on time series.

Value

A list is returned of class modl containing:

tserie

Original time serie.

tserieDF

Time serie converted to data frame.

method

Method used to build the model.

algorithm

If method is data mining, indicates wich algorithm was used.

horizon

Horizon for the splitting.

model

Model result from caret. It is a list, result of the caret::train function.

errors

Contains three different metrics to evaluate the model.

Author(s)

Alberto Vico Moreno

References

http://topepo.github.io/caret/index.html

See Also

prep modl.arima, modl.tsToDataFrame, modl.trControl, modl.dataMining

Examples

p <- prep(AirPassengers)
modl(p,method='arima')
modl(p,method='dataMining',algorithm='rpart')