custom-distance-metrics

Custom Distance Metrics for Scikit-Learn

When clustering spatial data, you often want to weigh dimensions differently or apply a non-standard metric.

Usage

You can implement custom metrics in sklearn or scipy. Scikit-learn's DBSCAN allows a custom distance metric passed as a callable.

Example:

import numpy as np
from sklearn.cluster import DBSCAN

def custom_metric(x, y, weight=1.0):
    dx = x[0] - y[0]
    dy = x[1] - y[1]
    return np.sqrt((weight * dx)**2 + ((2 - weight) * dy)**2)

# Usage in DBSCAN
dbscan = DBSCAN(eps=5, min_samples=3, metric=custom_metric, metric_params={'weight': 1.5})
clusters = dbscan.fit_predict(data)

Alternatively, for performance, precompute the distance matrix or use cdist/pdist with custom metrics if needed, but the callable approach is simplest.