verticapy.machine_learning.memmodel.cluster.BisectingKMeans¶

class verticapy.machine_learning.memmodel.cluster.BisectingKMeans(clusters: Annotated[list | ndarray, 'Array Like Structure'], children_left: Annotated[list | ndarray, 'Array Like Structure'], children_right: Annotated[list | ndarray, 'Array Like Structure'], cluster_size: Annotated[list | ndarray, 'Array Like Structure'] | None = None, cluster_score: Annotated[list | ndarray, 'Array Like Structure'] | None = None, p: int = 2)¶

InMemoryModel implementation of BisectingKMeans.

Parameters¶

clusters: ArrayLike: list of the model’s cluster centers.
children_left: ArrayLike: A list of node IDs, where children_left[i] is the node ID of the left child of node i.
children_right: ArrayLike: A list of node IDs, where children_right[i] is the node ID of the right child of node i.
cluster_size: ArrayLike: A list of sizes, where cluster_size[i] is the number of elements in node i.
cluster_score: ArrayLike: A list of scores, where cluster_score[i] is the score for internal node i. The score is the ratio between the within -cluster sum of squares of the node and the total within-cluster sum of squares.
p: int, optional: The p corresponding to one of the p-distances.

Attributes¶

Attributes are identical to the input parameters, followed by an underscore (‘_’).

Examples¶

Initalization

Import the required module.

from verticapy.machine_learning.memmodel.cluster import BisectingKMeans

A BisectingKMeans model is defined by its clusters centroids, left and right child node id’s of given node. In this example, we will use the following:

clusters = [
    [0.5, 0.6],
    [1, 2],
    [100, 200],
    [10, 700],
    [-100, -200],
]


children_left = [1, 3, None, None, None]

children_right = [2, 4, None, None, None]

Let’s create a BisectingKMeans model.

model_bkm = BisectingKMeans(clusters, children_left, children_right)

Create a dataset.

data = [[2, 3]]

Making In-Memory Predictions

Use predict() method to do predictions.

model_bkm.predict(data)[0]
Out[7]: 4

Use predict_proba() method to compute the predicted probabilities for each cluster.

model_bkm.predict_proba(data)
Out[8]: array([[0.32996436, 0.66034105, 0.00424426, 0.00133974, 0.00411059]])

Use transform() method to compute the distance from each cluster.

model_bkm.transform(data)
Out[9]: 
array([[  2.83019434,   1.41421356, 220.02954347, 697.04590954,
        227.18494668]])

Use to_graphviz() method to generate code for a Graphviz tree.

model_bkm.to_graphviz()
Out[10]: 'digraph Tree {\ngraph [rankdir = "LR"];\n0 [label="0", shape="none"]\n0 -> 1 [label=""]\n0 -> 2 [label=""]\n1 [label="1", shape="none"]\n1 -> 3 [label=""]\n1 -> 4 [label=""]\n2 [label="2", shape="none"]\n3 [label="3", shape="none"]\n4 [label="4", shape="none"]\n}'

Use plot_tree() method to draw the input tree.

model_bkm.plot_tree()

../_images/machine_learning_cluster_bisecting_kmeans.png

Note

plot_tree() requires the Graphviz module.

Deploy SQL Code

Let’s use the following column names:

cnames = ['col1', 'col2']

Use predict_sql() method to get the SQL code needed to deploy the model using its attributes.

model_bkm.predict_sql(cnames)
Out[12]: '(CASE WHEN col1 IS NULL OR col2 IS NULL THEN NULL ELSE (CASE WHEN POWER(POWER(col1 - 1.0, 2) + POWER(col2 - 2.0, 2), 1/2) < POWER(POWER(col1 - 100.0, 2) + POWER(col2 - 200.0, 2), 1/2) THEN (CASE WHEN POWER(POWER(col1 - 10.0, 2) + POWER(col2 - 700.0, 2), 1/2) < POWER(POWER(col1 - -100.0, 2) + POWER(col2 - -200.0, 2), 1/2) THEN 3 ELSE 4 END) ELSE 2 END) END)'

Use predict_proba_sql() method to get the SQL code needed to deploy the model that computes predicted probabilities.

model_bkm.predict_proba_sql(cnames)
Out[13]: 
['(CASE WHEN POWER(POWER(col1 - 0.5, 2) + POWER(col2 - 0.6, 2), 1 / 2) = 0 THEN 1.0 ELSE 1 / (POWER(POWER(col1 - 0.5, 2) + POWER(col2 - 0.6, 2), 1 / 2)) / (1 / (POWER(POWER(col1 - 0.5, 2) + POWER(col2 - 0.6, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 1.0, 2) + POWER(col2 - 2.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 100.0, 2) + POWER(col2 - 200.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 10.0, 2) + POWER(col2 - 700.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - -100.0, 2) + POWER(col2 - -200.0, 2), 1 / 2))) END)',
 '(CASE WHEN POWER(POWER(col1 - 1.0, 2) + POWER(col2 - 2.0, 2), 1 / 2) = 0 THEN 1.0 ELSE 1 / (POWER(POWER(col1 - 1.0, 2) + POWER(col2 - 2.0, 2), 1 / 2)) / (1 / (POWER(POWER(col1 - 0.5, 2) + POWER(col2 - 0.6, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 1.0, 2) + POWER(col2 - 2.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 100.0, 2) + POWER(col2 - 200.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 10.0, 2) + POWER(col2 - 700.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - -100.0, 2) + POWER(col2 - -200.0, 2), 1 / 2))) END)',
 '(CASE WHEN POWER(POWER(col1 - 100.0, 2) + POWER(col2 - 200.0, 2), 1 / 2) = 0 THEN 1.0 ELSE 1 / (POWER(POWER(col1 - 100.0, 2) + POWER(col2 - 200.0, 2), 1 / 2)) / (1 / (POWER(POWER(col1 - 0.5, 2) + POWER(col2 - 0.6, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 1.0, 2) + POWER(col2 - 2.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 100.0, 2) + POWER(col2 - 200.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 10.0, 2) + POWER(col2 - 700.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - -100.0, 2) + POWER(col2 - -200.0, 2), 1 / 2))) END)',
 '(CASE WHEN POWER(POWER(col1 - 10.0, 2) + POWER(col2 - 700.0, 2), 1 / 2) = 0 THEN 1.0 ELSE 1 / (POWER(POWER(col1 - 10.0, 2) + POWER(col2 - 700.0, 2), 1 / 2)) / (1 / (POWER(POWER(col1 - 0.5, 2) + POWER(col2 - 0.6, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 1.0, 2) + POWER(col2 - 2.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 100.0, 2) + POWER(col2 - 200.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 10.0, 2) + POWER(col2 - 700.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - -100.0, 2) + POWER(col2 - -200.0, 2), 1 / 2))) END)',
 '(CASE WHEN POWER(POWER(col1 - -100.0, 2) + POWER(col2 - -200.0, 2), 1 / 2) = 0 THEN 1.0 ELSE 1 / (POWER(POWER(col1 - -100.0, 2) + POWER(col2 - -200.0, 2), 1 / 2)) / (1 / (POWER(POWER(col1 - 0.5, 2) + POWER(col2 - 0.6, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 1.0, 2) + POWER(col2 - 2.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 100.0, 2) + POWER(col2 - 200.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - 10.0, 2) + POWER(col2 - 700.0, 2), 1 / 2)) + 1 / (POWER(POWER(col1 - -100.0, 2) + POWER(col2 - -200.0, 2), 1 / 2))) END)']

Use transform_sql() method to get the SQL code needed to deploy the model that computes distance from each cluster.

model_bkm.transform_sql(cnames)
Out[14]: 
['POWER(POWER(col1 - 0.5, 2) + POWER(col2 - 0.6, 2), 1 / 2)',
 'POWER(POWER(col1 - 1.0, 2) + POWER(col2 - 2.0, 2), 1 / 2)',
 'POWER(POWER(col1 - 100.0, 2) + POWER(col2 - 200.0, 2), 1 / 2)',
 'POWER(POWER(col1 - 10.0, 2) + POWER(col2 - 700.0, 2), 1 / 2)',
 'POWER(POWER(col1 - -100.0, 2) + POWER(col2 - -200.0, 2), 1 / 2)']

Hint

This object can be pickled and used in any in-memory environment, just like SKLEARN models.

__init__(clusters: Annotated[list | ndarray, 'Array Like Structure'], children_left: Annotated[list | ndarray, 'Array Like Structure'], children_right: Annotated[list | ndarray, 'Array Like Structure'], cluster_size: Annotated[list | ndarray, 'Array Like Structure'] | None = None, cluster_score: Annotated[list | ndarray, 'Array Like Structure'] | None = None, p: int = 2) → None¶

Methods

`__init__`(clusters, children_left, children_right)
`get_attributes`()	Returns the model attributes.
`plot_tree`([pic_path])	Draws the input tree.
`predict`(X)	Predicts using the `BisectingKMeans` model.
`predict_proba`(X)	Predicts the probability of each input to belong to the model clusters.
`predict_proba_sql`(X)	Returns the SQL code needed to deploy the model probabilities.
`predict_sql`(X)	Returns the SQL code needed to deploy the `BisectingKMeans` model using its attributes.
`set_attributes`(**kwargs)	Sets the model attributes.
`to_graphviz`([round_score, percent, ...])	Returns the code for a Graphviz tree.
`transform`(X)	Transforms and returns the distance to each cluster.
`transform_sql`(X)	Transforms and returns the SQL distance to each cluster.

Attributes

object_type

Must be overridden in child class