Weighted KMeans Clustering for Geolocational Problem

Repo for weighted k means clustering for specifically geo locational problems.

For an example and mathematical explanation:

https://emrahcimren.github.io/data%20science/Greenfield-Analysis-with-Weighted-Clustering/

Prerequisites

Install environment.yml for prerequisites.

conda env create -f environment.yml

To recreate environment.yml

conda env export > environment.yml

To create requirements.txt from environment.yml

pip freeze > requirements.txt

Installation

pip install cimren-wkmeans-geo

Inputs

input_locations is a pandas dataframe with the following format.

number_of_clusters: Number of clusters to be created

minimum_elements_in_a_cluster: Minimum elements in a cluster

maximum_elements_in_a_cluster: Maximum elements in a cluster

maximum_volume_in_a_cluster: Maximum volume that can fit in a cluster; if set to None, then it is disabled

maximum_iteration: How many maximum number of steps the algorithm takes to stop if it does not find the solution

enable_minimum_maximum_elements_in_a_cluster: True/False to enable minimum and maximum cluster size

objective_range: Acceptable difference between objectives at each iteration

Data

Package has a sample data set

from wkmeans_geo.src import data
data.locations_test
data.number_of_clusters
data.minimum_elements_in_a_cluster
data.maximum_elements_in_a_cluster
data.maximum_volume_in_a_cluster
data.maximum_iteration
data.enable_minimum_maximum_elements_in_a_cluster
data.objective_range

How to use

from wkmeans_geo.src import data
from wkmeans_geo import wkmeans_clustering as wkc
clusters, locations_with_clusters = wkc.calculate_clusters(
                                       data.locations_test,
                                       data.number_of_clusters,
                                       data.minimum_elements_in_a_cluster,
                                       data.maximum_elements_in_a_cluster,
                                       data.maximum_volume_in_a_cluster,
                                       data.maximum_iteration,
                                       data.objective_range,
                                       data.enable_minimum_maximum_elements_in_a_cluster)