search

config-i1 / smooth

The set of functions used for time series analysis and in forecasting.
89 stars 20 forks source link

Introduce parameter in functions for cumulative forecasts #77

Closed config-i1 closed 7 years ago

config-i1 commented 7 years ago

This means that, for example, es() would return only one aggregated point forecast for the lead time h and similar things for upper bound of prediction interval. This then can be used for ordering policy etc.

A connected thing - cost function based on cumulative forecasts. Should be used if we are interested in inventory control, as it decreases safety stock.

config-i1 commented 7 years ago

To do:

  1. Cumulative prediction intervals for multiplicative models (non-parametric and semiparametric at least);
  2. How to represent the thing on graph.
  3. Error measures for cumulative forecasts.
config-i1 commented 7 years ago
  1. Not really possible to do because of the contradiction between the multiplicative errors and the idea of cumulative forecasts. Currently simulation is used for all types of intervals here.
  2. graphmaker() is updated in 3f8643e971f0771a2df334bb5a2e7b48672f302e
config-i1 commented 7 years ago
  1. Sort of done in 6fc3618f3d3cc1c8fbd8f094063dcd9a63142b1e. At least this is a fix of the problem.