dcpython/kprototypes.py at main · ganml/dcpython · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
# -*- coding: utf-8 -*-
"""
Created on Tue Jan 28 20:22:56 2025

@author: gjgan
"""

import numpy as np
import matplotlib.pyplot as plt
from sklearn.impute import SimpleImputer
from scipy import stats
from ucimlrepo import fetch_ucirepo
from dcutil import createCM

def estBeta(Xn, Xc):
    numVar = np.mean(np.var(Xn, axis=0))
    vv = np.zeros(Xc.shape[1])
    for j in range(Xc.shape[1]):
        nv, nc = np.unique(Xi[:,j], return_counts=True)
        vp = nc/np.sum(nc)
        vv[j] = 1 - np.sum(np.square(vp))
    return numVar / np.mean(vv)


def kproto(Xn, Xc, k=3, beta=None, tol=1e-8, maxit=100):
    Xn = np.ascontiguousarray(Xn)
    Xc = np.ascontiguousarray(Xc)
    n, d1 = Xn.shape
    nc, d2 = Xc.shape
    if n != nc:
       raise ValueError("dimension mismatch")
    if beta is None:
        beta = estBeta(Xn, Xc)
    ind = np.random.choice(n, k, replace=False)
    clusterCentersn = Xn[ind,:]
    clusterCentersc = Xc[ind,:]
    dm = np.zeros((n,k))
    for i in range(k):
        dm[:,i] = np.sum(np.square(Xn-clusterCentersn[i,:]), axis=1) + beta * np.count_nonzero(Xc-clusterCentersc[i,:], axis=1)
    clusterMembership = np.argmin(dm, axis=1)
    objectiveValue = np.sum(dm[list(range(n)), clusterMembership]).item()
    numIter = 1
    while numIter < maxit:
        # update cluster centers
        for i in range(k):
            bInd = clusterMembership==i
            if np.any(bInd):
                clusterCentersn[i,:] = np.mean(Xn[bInd], axis=0)
                clusterCentersc[i,:] = stats.mode(Xc[bInd])[0]
            else:
                clusterCentersn[i,:] = Xn[np.random.randint(0, n),:]
                clusterCentersc[i,:] = Xc[np.random.randint(0, n),:]
        # update cluster membership
        for i in range(k):
            dm[:,i] = np.sum(np.square(Xn-clusterCentersn[i,:]), axis=1) + beta * np.count_nonzero(Xc-clusterCentersc[i,:], axis=1)
        clusterMembership = np.argmin(dm, axis=1)
        objectiveValue_ = objectiveValue
        objectiveValue = np.sum(dm[list(range(n)), clusterMembership]).item()
        numIter += 1
        if np.abs(objectiveValue - objectiveValue_) < tol:
            break;
    return clusterMembership, clusterCentersn, clusterCentersc, objectiveValue, numIter

def kproto2(Xn, Xc, k=3, beta=None, numrun=10, maxit=100):
    bestCM, bestCCn, bestCCc, bestOV, bestIters = kproto(Xn, Xc, k=k, beta=beta, maxit=maxit)
    vOV = np.zeros(numrun)
    vOV[0] = bestOV
    for i in range(numrun-1):
        cp, ccn, ccc, ov, iters = kproto(Xn, Xc, k=k, beta=beta, maxit=maxit)
        vOV[i+1] =ov
        if ov < bestOV:
            bestCM, bestCCn, bestCCc, bestOV, bestIters = cp, ccn, ccc, ov, iters
    return bestCM, bestCCn, bestCCc, bestOV, bestIters, vOV

# examples

heart_disease = fetch_ucirepo(id=45)

X = heart_disease.data.features
y = heart_disease.data.targets.copy()
y[y>0] = 1

print(X.isnull().sum())

imp = SimpleImputer(missing_values=np.nan, strategy='most_frequent')
imp.fit(X)
Xi = imp.transform(X)

varNames = list(X.columns)
numNames = ['age', 'trestbps', 'chol', 'thalach', 'oldpeak', 'ca']
numInd = [varNames.index(s) for s in numNames]
catInd = [varNames.index(s) for s in set(varNames).difference(numNames)]
catInd.sort()

Xn = Xi[:, numInd]
Xc = Xi[:, catInd].astype(int)

vMin = np.min(Xn, axis=0)
vMax = np.max(Xn, axis=0)
Xn = (Xn-vMin)/(vMax-vMin)

print(np.min(Xn, axis=0))
print(np.max(Xn, axis=0))
print(np.min(Xc, axis=0))
print(np.max(Xc, axis=0))
print(estBeta(Xn, Xc))

# use default parameters
bcm, bccn, bccc, bov, biters, vOV = kproto2(Xn, Xc, k=2, numrun=100)
cm1 = createCM(y, bcm)
print(cm1)
print([bov, biters])

fig, ax = plt.subplots(1, 1, figsize=(6, 4))
ax.scatter(range(len(vOV)), vOV, color="black")
ax.set_xlabel("Run number")
ax.set_ylabel("Objective function value")
fig.savefig("heartov.pdf", bbox_inches='tight')

# use beta=1
bcm, bccn, bccc, bov, biters, vOV = kproto2(Xn, Xc, k=2, beta=1, numrun=100)
cm1 = createCM(y, bcm)
print(cm1)
print([bov, biters])

fig, ax = plt.subplots(1, 1, figsize=(6, 4))
ax.scatter(range(len(vOV)), vOV, color="black")
ax.set_xlabel("Run number")
ax.set_ylabel("Objective function value")
fig.savefig("heartov1.pdf", bbox_inches='tight')

# use beta=0
bcm, bccn, bccc, bov, biters, vOV = kproto2(Xn, Xc, k=2, beta=0, numrun=100)
cm1 = createCM(y, bcm)
print(cm1)
print([bov, biters])

fig, ax = plt.subplots(1, 1, figsize=(6, 4))
ax.scatter(range(len(vOV)), vOV, color="black")
ax.set_xlabel("Run number")
ax.set_ylabel("Objective function value")
fig.savefig("heartov1.pdf", bbox_inches='tight')