BisectingKMeans

In [ ]:
BisectingKMeans(name: str,
                n_cluster: int = 8,
                bisection_iterations: int = 1,
                split_method: str = 'sum_squares',
                min_divisible_cluster_size: int = 2,
                distance_method: str = "euclidean",
                init: str = "kmeanspp",
                max_iter: int = 300,
                tol: float = 1e-4)

Creates a bisecting k-means object by using Vertica function BISECTING_KMEANS on the data. K-means clustering is a method of vector quantization, originally from signal processing, that aims to partition n observations into k clusters, where each observation belongs to the cluster with the nearest mean (cluster centers or cluster centroid), serving as a prototype of the cluster. This results in a partitioning of the dataspace into Voronoi cells. Bisecting k-means combines k-means and hierarchical clustering.

Parameters

Name Type Optional Description
name
str
❌
 Name of the model to be stored in the database.
n_cluster
int
✓
 Number of clusters
bisection_iterations
int
✓
 The number of iterations the bisecting k-means algorithm performs for each bisection step. This corresponds to how many times a standalone k-means algorithm runs in each bisection step. A value greater than 1 allows the algorithm to run and choose the best k-means run within each bisection step. Note that if you are using the kmeanspp function, the bisection_iterations value is always 1 because kmeanspp is more both more reliable and expensive, so it only needs to be run once.
split_method
str
✓
 The method used to choose a cluster to bisect/split.
  • size : Choose the largest cluster to bisect.
  • sum_squares : Choose the cluster with the largest withinSS to bisect.
min_divisible_cluster_size
int
✓
 The minimum number of points of a divisible cluster. Must be greater than or equal to 2.
distance_method
str
✓
 The measure for distance between two data points. Only Euclidean distance is supported at this time.
init
str / list
✓
 The method to use to find the initial cluster centers.
  • kmeanspp : Uses the KMeans++ method to initialize the centers.
  • pseudo : Uses "pseudo center" approach used by Spark, bisects given center without iterating over points.
max_iter
int
✓
 The maximum number of iterations the algorithm performs.
tol
float
✓
 Determines whether the algorithm has converged. The algorithm is considered converged when no center has moved more than a distance of 'tol' from the previous iteration.

Attributes

After the object is created, all parameters become attributes. Additional attributes will be created when fitting the model:

Name Type Description
cluster_centers_
tablesample
 Clusters result of the algorithm.
metrics_
tablesample
 Metrics with which to evaluate the model.
input_relation
str
 The training relation.
X
list
 List of the predictors.

Methods

Name Description
deploySQL Returns the SQL code needed to deploy the model.
drop Drops the model from the Vertica database.
fit Trains the model.
get_attr Returns the model attribute.
get_params Returns the model parameters.
plot Draws the k-means clusters.
plot_tree Draws the input BKtree.
predict Predicts using the input relation.
set_params Sets the parameters of the model.
to_graphviz Returns the code for a Graphviz tree.
to_memmodel Converts a specified Vertica model to a memModel model.
to_python Returns the Python code needed to deploy the model without using built-in Vertica functions.
to_sql Returns the SQL code needed to deploy the model without using Vertica built-in functions.

Example

In [24]:
from verticapy.learn.cluster import BisectingKMeans
model = BisectingKMeans(name = "public.bkmeans_iris",
                        n_cluster = 8,
                        init = "kmeanspp",
                        max_iter = 300,
                        tol = 1e-4)
display(model)

<BisectingKMeans>