In tidyclust: A Common API to Clustering

#| child: aaa.Rmd
#| include: false

r descr_models("mean_shift", "meanShiftR")

Tuning Parameters

#| label: meanShiftR-param-info
#| echo: false
defaults <-
  tibble::tibble(
    tidyclust = c("bandwidth"),
    default = c("no default")
  )

param <-
  mean_shift() |>
  set_engine("meanShiftR") |>
  set_mode("partition") |>
  make_parameter_list(defaults)

This model has r nrow(param) tuning parameters:

#| label: meanShiftR-param-list
#| echo: false
#| results: asis
param$item

Translation from tidyclust to the original package (partition)

#| label: meanShiftR-cls
mean_shift(bandwidth = 0.5) |>
  set_engine("meanShiftR") |>
  set_mode("partition") |>
  translate_tidyclust()

Preprocessing requirements

#| child: template-makes-dummies.Rmd

Unlike the LPCM engine, meanShiftR::meanShift() does not scale variables internally and operates on the raw data scale. The bandwidth value is used directly as a per-dimension kernel width on the original variables, and a scalar bandwidth is recycled to a per-column vector. Because of this, appropriate bandwidths typically depend on the spread of the predictors. Standardizing predictors before fitting (for example, with [recipes::step_normalize()]) is recommended; otherwise the default dials::bandwidth() range of c(0.01, 1) may be too narrow.

What does it mean to predict?

To predict the cluster assignment for a new observation, the mean shift procedure is run from the new point against the training data's kernel density estimate. The observation is assigned to the cluster whose training mode is closest to the converged value by Euclidean distance.

References

Cheng, Y. (1995). Mean shift, mode seeking, and clustering. IEEE Transactions on Pattern Analysis and Machine Intelligence, 17(8), 790–799. doi: 10.1109/34.400568
Comaniciu, D., & Meer, P. (2002). Mean shift: A robust approach toward feature space analysis. IEEE Transactions on Pattern Analysis and Machine Intelligence, 24(5), 603–619. doi: 10.1109/34.1000236
Lisic, J. (2015). Parcel Level Agricultural Land Cover Prediction (Doctoral dissertation, George Mason University).