deepAR multivariado ejemplo

Gluon
1

Junto con saludar a la comunidad MXNET me gustaría aprender como es la sintaxis para datos multivariados al programar DEEPAR en gluonTS. Mi objetivo es predecir multiples series.

Otra consulta es la aclaración si es que los campos dynamic_feat y cat son para variables exogenas o regresoras independientes, pero la primera para datos continuos y la otra para datos categoricos.

Saludos, espero una pronta respuesta

2

Welcome to the MXNet Community!

DeepAR works with multivariate time series. For instance your input may look like the following:

{"start": "2009-11-01 00:00:00", "target": [4.3, "NaN", 5.1, ...], "cat": [0, 1], "dynamic_feat": [[1.1, 1.2, 0.5, ...]]}
{"start": "2012-01-30 00:00:00", "target": [1.0, -5.0, ...], "cat": [2, 3], "dynamic_feat": [[1.1, 2.05, ...]]}
{"start": "1999-01-30 00:00:00", "target": [2.0, 1.0], "cat": [1, 4], "dynamic_feat": [[1.3, 0.4]]}

Each dictionary contains information about a timeseries. You can use the entry cat to define your categorical variables. The catgeorical features need to be encoded as integer values. So in the above example each timeseries has 2 categorical features. You can use dynamic_feat to describe vectors of custom time series features.

1 Like
3

thank you for response!

Two queries:

  • What is the syntax for adding exogenous variables? dynamic_feat represents that?

  • It is assumed that when training the model, can I then make a prediction about the two series or will it always be made on just one variable?