gower | Gower 's distance for R | Architecture library

hclust returns just a dendrogram representing clusters inside clusters. The silhouette score is defined on one specific clustering and not on all possible clusterings. This is why it works with Partitioning Arround Medoids (PAM) out of the box. In hierarchical clustering, however, one needs to decide first which clustering to chose by cutting dendrogram tree.

This is how to plot the silhouettes for a hierarchical clustering using k=5 clusters:

Once you get to the script tag, nokogiri's work is done, and it's time to parse JSON.

This problem has been fixed in the developmental version of {parsnip} (>0.1.7). You can install it by running remotes::install_github("tidymodels/parsnip").

You may use maxsum=. Example:

@importFrom recipes prep bake had to be added to the .R file

Errors like this in Shiny apps are almost always a result of missing packages, which the logs confirm.

Turns out in this case I think the missing package is kernlab, which I only found by reading the documentation given here: https://topepo.github.io/caret/train-models-by-tag.html#Support_Vector_Machines. It's a suggested package, not imported, so the command suggested in the comments by heds1 would sort this out.

You definitely were passing the vector of variables correctly to the recipe -- no problem there!

You were running into other problems with your model fitting. An xgboost model requires all predictors to be numeric, so if you convert something like zip code to factors, you need to then use step_dummy(). If you have something of high cardinality like zip codes, you probably will need to handle new levels or unknown levels as well.

As touched upon in the help page, the core of hdbscan is 1) calculating the mutual reachability distance and 2) applying the single linkage algorithm. Since you do not have that many data points and your distance metric is pre-computed, you can see your clustering is decided by the single linkage:

In order to make procedures 4 - 7 work we needed to make the adjustments listed in the conclusions section of Using microbenchmark and ggplot2 to plot runtimes:

1. Wrap the original procedure in a function that we can use as the unit of analysis for microbenchmark(), and include a size argument
2. Modify the procedure to use size as a variable where necessary
3. Modify the procedure to access objects from previous steps, based on the size argument
4. Modify the procedure to write its outputs with assign() and size if these are needed for subsequent procedure steps

The modified code looks like this: