-
Notifications
You must be signed in to change notification settings - Fork 2
/
dslvq.py
197 lines (162 loc) · 7.06 KB
/
dslvq.py
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
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
# -*- coding: utf-8 -*-
"""
Created on Fri Sep 21 22:05:22 2018
@author: USER
"""
import numpy as np
import csv
#dataClass = [[[1.0,1.0,0.0,0.0],0],[[0.0,0.0,0.0,1.0],1]]
#dataTraining = [[[0.0,0.0,1.0,1.0],1],[[1.0,0.0,0.0,0.0],0],[[0.0,1.0,1.0,0.0],1]]
def read_csv(file_name):
array_2D = []
with open(file_name, 'rb') as csvfile:
read = csv.reader(csvfile, delimiter=';')
for row in read:
array_2D.append(row)
return array_2D
data1 = read_csv('data/dataColor.csv')
data2 = read_csv('data/dataTrainColor.csv')
data3 = read_csv('data/dataTestColor.csv')
dataTrain = ((np.array(data1[:]))[:,1:-1]).astype(np.float16).tolist()
dataC = ((np.array(data2[:]))[:,1:-1]).astype(np.float16).tolist()
dataTesting = ((np.array(data3[:]))[:,1:-1]).astype(np.float16).tolist()
classDataTrain = ((np.array(data1[:]))[:,-1:]).astype(int).tolist()
classDataClass = ((np.array(data2[:]))[:,-1:]).astype(int).tolist()
dataTraining = []
dataClass = []
#dataTesting = []
for i in range(len(dataTrain)):
dataArray = []
dataArray.append(dataTrain[i])
dataArray.append(classDataTrain[i][0])
dataTraining.append(dataArray)
for i in range(len(dataC)):
dataArray2 = []
dataArray2.append(dataC[i])
dataArray2.append(classDataClass[i][0])
dataClass.append(dataArray2)
alpha = 0.1
beta = 0.1*alpha
epsilon= 0.35
epsilon2= 0.2
# Initialize Weight Matrix
weightMatrix = np.zeros((len(dataClass),len(dataTraining[0][0])),dtype=np.float64)
for i in range(len(weightMatrix)):
for j in range(len(dataTraining[i][0])):
weightMatrix[i][j] = dataClass[i][0][j]
def takeFirst(elem):
return elem[0]
def norm(listElement):
normResult = listElement[:]
sumValue = sum(listElement)
for i in range(len(normResult)):
normResult[i] = np.divide(listElement[i],sumValue)
return normResult
def getWN(diff):
return norm(diff)
def threshold(i):
if (i < 0.0001):
return 0.0001
elif (i > 1):
return 1
else:
return i
countLVQ1 = 0
countLVQ2 = 0
countLVQ21 = 0
countLVQ3 = 0
wFeature = [1]*len(dataClass[0][0])
wF = [1]*len(dataClass[0][0])
# ITERATION LVQ
iterasi = 50
for x in range(iterasi):
# Find Manhattan Distance
jarak = np.zeros((len(dataClass),len(dataTraining)),dtype=np.float64)
for i in range(len(dataTraining)):
for k in range(len(dataClass)):
jarak[k][i] = 0
for j in range(len(dataTraining[i][0])):
jarak[k][i] += wFeature[j] * np.abs(dataTraining[i][0][j] - weightMatrix[k][j])
jMatrix = [[]] * len(dataTraining)
for i in range(len(dataTraining)):
jVal = []
for j in range(len(dataClass)):
jVal.append([jarak[j][i],j])
jMatrix[i] = jVal
jMatrix[i].sort(key=takeFirst)
'''
# Initialize jValue
jValue = np.zeros((len(dataTraining)),dtype=int)
for i in range(len(jValue)):
winnerClass = dataClass[0][1]
minValue = 999999
for j in range(len(jarak)):
if(jarak[j][i] < minValue):
minValue = jarak[j][i]
winnerClass = j
jValue[i] = winnerClass
'''
for i in range(len(dataTraining)):
for j in range(len(weightMatrix)):
t = dataTraining[i][1]
yc1 = int(jMatrix[i][0][1]) # The nearest class from the data training
yc2 = int(jMatrix[i][1][1]) # The second nearest class from the data training
dc1 = jMatrix[i][0][0] # The distance of the nearest class from the data training
dc2 = jMatrix[i][1][0] # The distance of the second nearest class from the data training
di = [np.abs(a-b) for a,b in zip(dataTraining[i][0],dataClass[yc1][0])]
dj = [np.abs(a-b) for a,b in zip(dataTraining[i][0],dataClass[yc2][0])]
wn = getWN([a-b for a,b in zip(di,dj)])
for k in range(len(weightMatrix[j])):
if ( (min(np.divide(dc1,dc2),np.divide(dc2,dc1)) > ((1-epsilon2)*(1+epsilon2))) and (yc1 == t and yc2 == t) ):
countLVQ3 += 1
if(j == yc1 or j == yc2):
weightMatrix[j][k] = (1-beta)*weightMatrix[j][k] + beta*dataTraining[i][0][k]
elif ( (min(np.divide(dc1,dc2),np.divide(dc2,dc1)) > (1-epsilon)) and (max(np.divide(dc1,dc2),np.divide(dc2,dc1)) < (1+epsilon)) and ((yc1 == t and yc2 != t) or (yc1 != t and yc2 == t)) ):
countLVQ21 += 1
if(j == yc1):
if(t == yc1):
weightMatrix[j][k] = (1-alpha)*weightMatrix[j][k] + alpha*dataTraining[i][0][k]
else:
weightMatrix[j][k] = (1+alpha)*weightMatrix[j][k] - alpha*dataTraining[i][0][k]
elif(j == yc2):
if(t == yc2):
weightMatrix[j][k] = (1-alpha)*weightMatrix[j][k] + alpha*dataTraining[i][0][k]
else:
weightMatrix[j][k] = (1+alpha)*weightMatrix[j][k] - alpha*dataTraining[i][0][k]
elif ((np.divide(jMatrix[i][0][0],jMatrix[i][1][0]) > (1 - epsilon)) and (np.divide(jMatrix[i][1][0],jMatrix[i][0][0]) < (1 + epsilon)) and (jMatrix[i][1][1] == dataTraining[i][1])):
countLVQ2 += 1
if(j == jMatrix[i][0][1]):
weightMatrix[j][k] = (1+alpha)*weightMatrix[j][k] + alpha*dataTraining[i][0][k]
elif(j == jMatrix[i][1][1]):
weightMatrix[j][k] = (1-alpha)*weightMatrix[j][k] + alpha*dataTraining[i][0][k]
elif (j == jMatrix[i][0][1]):
countLVQ1 += 1
if(j == dataTraining[i][1]):
weightMatrix[j][k] = (1-alpha)*weightMatrix[j][k] + alpha*dataTraining[i][0][k]
else:
weightMatrix[j][k] = (1+alpha)*weightMatrix[j][k] - alpha*dataTraining[i][0][k]
#if ( (min(np.divide(dc1,dc2),np.divide(dc2,dc1)) > ((1-epsilon2)*(1+epsilon2))) and ((yc1 == t and yc2 != t) or (yc1 != t and yc2 == t)) ):
if ( ((yc1 == t and yc2 != t) or (yc1 != t and yc2 == t)) ):
wFeature = norm([threshold( (1-alpha)*z + alpha*zw ) for z,zw in zip(wF,wn)])
alpha = 0.8*alpha
# Validation
#dataTest = [0.0,0.0,1.0,1.0]
#dataTest = [1.0,0.0,0.0,0.0]
#dataTest = [0.0,1.0,1.0,0.0]
dataTest = dataTesting[2]
classResult = 0
minValue = 99999
for i in range(len(weightMatrix)):
sumValue = 0
for j in range(len(weightMatrix[i])):
sumValue += np.power(dataTest[j] - weightMatrix[i][j],2)
if (sumValue < minValue):
minValue = sumValue
classResult = i
print('Class Result = '+str(classResult))
print(countLVQ1/4)
print(countLVQ2/4)
print(countLVQ21/4)
print(countLVQ3/4)
#res = [np.abs(a-b) for a,b in zip(dataClass[0][0], dataClass[1][0])]
#print(norm(res))