modeltime | Modeltime unlocks time series forecast models and machine | Machine Learning library

I use a custom step function that I have built for a package of mine that works just fine, it is in a local copy of the package (Not yet submitted to CRAN) link to function here: step_hai_fourier

Here is the session info (we can see healthyR.ai 0.0.2.9000 is loaded):

I want to extract, .metrics (RMSE) from a Rolling origin forecast resampling (tibble: 52 x 5) by "id" columns which consist of slices.

The replicating codes are given below. Here is my attempt.

Currently I am doing some experiments with hyperparameter tuning for XGBoost regression on time series, using a latin hypercube sampling strategy. When running the code below, all the models fail during the tune_grid operation. The cause seems to be the recipe object. I used step_dummy() to transform the value column of my univariate time series In the .notes object appears the Error message: preprocessor 1/1: Error: unused argument (values)

I found some other post where this issue popped up, but none of the solutions helped in my case.

as I come from a classical time series analysis approach, I am still kinda new to parameter tuning. As tuning all local models (couple of hundreds of time series for product demand in my case) turns out to be not even near scalability, I want to analyze first the effect of tuning time series with low accuracy values, to evaluate the trade-off between scalability and accuracy, to see if tuning is justified for a particular time series issue. When I run the code below, it seems like I didn't properly specify the ranges for the regular grid. I think so, because it seems odd, to retrieve only three combinations for a tree value with range from 50 to 2000. Is this standard behavior? Does changing the levels argument help in this case somehow? It didn't change anything in my case. Also, is there a way to retrieve the optimal number of folds for resampling, rather than guessing it? I hope for some advise or useful examples.

Thanks in advance!

I've got a simple CSV log file, reading it into a list and charting it is really easy in LINQPad, but I'd like to update the format and the ticks on the x-axis.

Here's my code:

below I created some fake forecast data using the tidyverse modeltime packages. I have got monthly data from 2016 and want to produce a test fc for 2020. As you can see, the data I load comes in wide format. For usage in modeltime I transform it to long data. After the modeling phase, I want to create a dataframe for the 2020 prediction values. For this purpose I need to somehow "unmelt" the data. In this process I am unfortunately losing a lot of variables. From 240 variables that I want to forecast I get only 49 in the end result. Maybe I am blind, or I do not know how to configure the modeltime functions correctly. I would really much appreciate some help. Thanks in advance!

When working with time series through the modeltime / tidymodels framework, the following workflow is presented to verify the performance of the models with cross validation.

I have publicly shares a small dataset on Google Drive and I have made the file accessible to anyone with the link.

I wish to download this file into R for analysis but I am having difficulty with unzipping the file from the temp directory.

My code looks as follows: