How can I plot a k-dsitance graph using python

Question

Im trying to plot the distance graph for a given value of min-points. I'm specifically looking for the
knee and corresponding epsilon values. I've been trying to use sklearn for my cause, but I can't seem to
find any function that returns these specific values. Can someone help me?

Answer 1

Hi there, instead of sklearn you could use this function to get what you want

import numpy as np
import pandas as pd
import math

def k_distances(X, n=None, dist_func=None):
    """Function to return array of k_distances.

    X - DataFrame matrix with observations
    n - number of neighbors that are included in returned distances (default number of attributes + 1)
    dist_func - function to count distance between observations in X (default euclidean function)
    """
    if type(X) is pd.DataFrame:
        X = X.values
    k=0
    if n == None:
        k=X.shape[1]+2
    else:
        k=n+1

    if dist_func == None:
        # euclidean distance square root of sum of squares of differences between attributes
        dist_func = lambda x, y: math.sqrt(
            np.sum(
                np.power(x-y, np.repeat(2,x.size))
            )
        )

    Distances = pd.DataFrame({
        "i": [i//10 for i in range(0, len(X)*len(X))],
        "j": [i%10 for i in range(0, len(X)*len(X))],
        "d": [dist_func(x,y) for x in X for y in X]
    })
    return np.sort([g[1].iloc[k].d for g in iter(Distances.groupby(by="i"))])

    X should be pandas.DataFrame or numpy.ndarray. n is number of neighbors that are in d-neighborhood. You should know this number. By default is number of attributes + 1.

To plot the distance using python use matplotlib

import matplotlib.pyplot as plt

d = k_distances(X,n,dist_func)
plt.plot(d)
plt.ylabel("k-distances")
plt.grid(True)
plt.show()

Answer 2

You probably want to use the matrix operations provided by numpy to speed up your distance matrix calculation.

def k_distances2(x, k):
    dim0 = x.shape[0]
    dim1 = x.shape[1]
    p=-2*x.dot(x.T)+np.sum(x**2, axis=1).T+ np.repeat(np.sum(x**2, axis=1),dim0,axis=0).reshape(dim0,dim0)
    p = np.sqrt(p)
    p.sort(axis=1)
    p=p[:,:k]
    pm= p.flatten()
    pm= np.sort(pm)
    return p, pm
m, m2= k_distances2(X, 2)
plt.plot(m2)
plt.ylabel("k-distances")
plt.grid(True)
plt.show()

Comments

I want to  measure the distance
 of the  object  in graphically  using  raspberry  pi