Skip to main content

Implementation of state-of-the-art distance metrics from research papers which can handle mixed-type data and missing values.

Project description


Implementation of state-of-the-art distance metrics from research papers which can handle mixed-type data and missing values. At the moment, HEOM, HVDM and VDM are tested and working. VDM and HVDM has been released recently so please report bugs, if there are any. Please feel free to help and contribute to the project as there is a lack of existing implementations of hetergeneous distance metrics.


Clone the repository with git clone. Install the necessary packages with pipenv install

Example - HEOM

# Example code of how the HEOM metric can be used together with Scikit-Learn
import numpy as np
from sklearn.neighbors import NearestNeighbors
from sklearn.datasets import load_boston
# Importing a custom metric class
from HEOM import HEOM

# Load the dataset from sklearn
boston = load_boston()
boston_data = boston["data"]
# Categorical variables in the data
categorical_ix = [3, 8]
# The problem here is that NearestNeighbors can't handle np.nan
# So we have to set up the NaN equivalent
nan_eqv = 12345

# Introduce some missingness to the data for the purpose of the example
row_cnt, col_cnt = boston_data.shape
for i in range(row_cnt):
    for j in range(col_cnt):
        rand_val = np.random.randint(20, size=1)
        if rand_val == 10:
            boston_data[i, j] = nan_eqv

# Declare the HEOM with a correct NaN equivalent value
heom_metric = HEOM(boston_data, categorical_ix, nan_equivalents = [nan_eqv])

# Declare NearestNeighbor and link the metric
neighbor = NearestNeighbors(metric = heom_metric.heom)

# Fit the model which uses the custom distance metric

# Return 5-Nearest Neighbors to the 1st instance (row 1)
result = neighbor.kneighbors(boston_data[0].reshape(1, -1), n_neighbors = 5)

Research Papers

The code have implemented based on the following literature: HEOM, VDM and HVDM:

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for distython, version 0.0.3
Filename, size File type Python version Upload date Hashes
Filename, size distython-0.0.3-py3-none-any.whl (8.0 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size distython-0.0.3.tar.gz (4.3 kB) File type Source Python version None Upload date Hashes View

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring DigiCert DigiCert EV certificate Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page