simple_kmeans_db: Simple kmeans routine that works in-database

In edgararuiz/modeldb: Fits Models Inside the Database

Simple kmeans routine that works in-database

Description

It uses 'tidyeval' and 'dplyr' to run multiple cycles of kmean calculations, expressed in dplyr formulas until an the optimal centers are found.

Usage

simple_kmeans_db(
  df,
  ...,
  centers = 3,
  max_repeats = 100,
  initial_kmeans = NULL,
  safeguard_file = "kmeans.csv",
  verbose = TRUE
)

Arguments

df	A Local or remote data frame
...	A list of variables to be used in the kmeans algorithm
centers	The number of centers. Defaults to 3.
max_repeats	The maximum number of cycles to run. Defaults to 100.
initial_kmeans	A local dataframe with initial centroid values. Defaults to NULL.
safeguard_file	Each cycle will update a file specified in this argument with the current centers. Defaults to 'kmeans.csv'. Pass NULL if no file is desired.
verbose	Indicates if the progress bar will be displayed during the model's fitting.

Details

Because each cycle is an independent 'dplyr' operation, or SQL operation if using a remote source, the latest centroid data frame is saved to the parent environment in case the process needs to be canceled and then restarted at a later point. Passing the current_kmeans as the initial_kmeans will allow the operation to pick up where it left off.

Examples

library(dplyr)

mtcars %>%
  simple_kmeans_db(mpg, qsec, wt) %>%
  glimpse()

edgararuiz/modeldb documentation built on Jan. 18, 2024, 4:53 a.m.