-
Notifications
You must be signed in to change notification settings - Fork 1
/
OBHSA.py
162 lines (137 loc) · 6.79 KB
/
OBHSA.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
import numpy as np
import pandas as pd
import sklearn
from sklearn import datasets,svm,metrics
from sklearn.model_selection import KFold
from sklearn.neighbors import KNeighborsClassifier
from sklearn.neural_network import MLPClassifier
from sklearn.model_selection import train_test_split
from sklearn.metrics import *
def reduce_features(solution, features):
selected_elements_indices = np.where(solution ==1)[0]
reduced_features = features[:, selected_elements_indices]
return reduced_features
def classification_accuracy(labels, predictions):
correct = np.where(labels == predictions)[0]
accuracy = correct.shape[0]/labels.shape[0]
return accuracy
def cal_pop_fitness(pop, features, labels, train_indices,val_indices,classifier):
test_accuracies = np.zeros(pop.shape[0])
val_accuracies = np.zeros(pop.shape[0])
idx = 0
val_pop_pred = np.zeros(shape=(pop.shape[0],val_indices.shape[0]))
for i,curr_solution in enumerate(pop):
reduced_features = reduce_features(curr_solution, features)
train_data = reduced_features[train_indices, :]
val_data=reduced_features[val_indices,:]
train_labels = labels[train_indices]
val_labels=labels[val_indices]
if classifier=='SVM':
SV_classifier = sklearn.svm.SVC(kernel='rbf',gamma='scale',C=5000)
SV_classifier.fit(X=train_data, y=train_labels)
val_predictions = SV_classifier.predict(val_data)
val_accuracies[idx] = classification_accuracy(val_labels, val_predictions)
idx = idx + 1
elif classifier == 'KNN':
knn=KNeighborsClassifier(n_neighbors=8)
knn.fit(train_data,train_labels)
val_predictions=knn.predict(val_data)
val_accuracies[idx]=classification_accuracy(val_labels,predictions)
idx = idx + 1
else :
mlp = MLPClassifier()
mlp.fit(train_data,train_labels)
val_predictions=mlp.predict(val_data)
val_accuracies[idx]=classification_accuracy(val_labels,predictions)
idx = idx + 1
val_pop_pred[i] = val_predictions
return val_accuracies,val_pop_pred
def get_vector(labels,pop_preds):
vector = np.zeros(shape=(pop_preds.shape[0],4))
for i in range(pop_preds.shape[0]):
preds = pop_preds[i]
acc = classification_accuracy(labels,preds)
pre = precision_score(labels,preds,average="macro")
rec = recall_score(labels,preds,average="macro")
f1 = f1_score(labels,preds,average="macro")
vector[i] = np.array([acc,pre,rec,f1])
return vector
def OBHSA(data_inputs,data_outputs,
popSize=20,
HMCR=0.9,
PAR=0.35,
classifier="SVM",
num_generations = 10 #Number of generations in each fold
):
print("\nOPPOSITION-BASED HARMONY SEARCH:\n")
population_output = np.zeros(shape = (popSize,data_inputs.shape[1],5))
vector_output = np.zeros(shape=(popSize,4,5)) #4 because acc,pre,rec,f1; and 5 because folds=5
num_samples = data_inputs.shape[0]
num_feature_elements = data_inputs.shape[1]
HM_shape=(popSize,num_feature_elements)
harmony_memory=np.random.randint(low=0,high=2,size=HM_shape)
NCHV = np.ones((1, num_feature_elements))
best_outputs = []
best_opp_outputs = []
kf=KFold(5,True,random_state=1)
fold=0
for train_indices,test_val_indices in kf.split(data_inputs):
print("Fold : ",fold+1)
val_indices,test_indices=train_test_split(test_val_indices,test_size=0.5,shuffle=True,random_state=8)
best_test_outputs=[]
harmony_memory=np.random.randint(low=0,high=2,size=HM_shape)
opposite_memory=1-harmony_memory
total_memory=np.concatenate((harmony_memory,opposite_memory),axis=0)
total_fitness,_ = cal_pop_fitness(total_memory,data_inputs,data_outputs,train_indices,val_indices,classifier)
fit_ind = np.argpartition(total_fitness, -popSize)[-popSize:]
harmony_memory=total_memory[fit_ind,:]
gen_fit = np.array([-1])
for currentIteration in range(num_generations):
NCHV = np.ones((1, num_feature_elements))
print("Generation : ", currentIteration+1)
fitness,val_pop_preds=cal_pop_fitness(harmony_memory,data_inputs,data_outputs,train_indices,val_indices,classifier)
best_outputs.append(np.max(fitness))
print("Best validation result : ", max(best_outputs))
if max(fitness)>max(gen_fit):
gen_fit = fitness
gen_labels = data_outputs[val_indices]
gen_preds = val_pop_preds
for i in range(num_feature_elements):
ran = np.random.rand()
if ran < HMCR:
index = np.random.randint(0, popSize)
NCHV[0, i] = harmony_memory[index, i]
pvbran = np.random.rand()
if pvbran < PAR:
pvbran1 = np.random.rand()
result = NCHV[0, i]
if pvbran1 < 0.5:
result =1-result
else:
NCHV[0, i] = np.random.randint(low=0,high=2,size=1)
new_fitness,_ = cal_pop_fitness(NCHV, data_inputs, data_outputs, train_indices, val_indices,classifier)
if new_fitness > min(fitness):
min_fit_idx = np.where(fitness == min(fitness))
harmony_memory[min_fit_idx, :] = NCHV
fitness[min_fit_idx] = new_fitness
opp_NCHV=1-NCHV
new_opp_fitness,_=cal_pop_fitness(opp_NCHV,data_inputs, data_outputs, train_indices, val_indices,classifier)
if new_opp_fitness > min(fitness):
min_fit_idx = np.where(fitness == min(fitness))
harmony_memory[min_fit_idx, :] = opp_NCHV
fitness[min_fit_idx] = new_opp_fitness
fitness,_ = cal_pop_fitness(harmony_memory, data_inputs, data_outputs, train_indices,val_indices,classifier)
best_match_idx = np.where(fitness == np.max(fitness))[0]
best_match_idx = best_match_idx[0]
best_solution = harmony_memory[best_match_idx, :]
best_solution_indices = np.where(best_solution == 1)[0]
best_solution_num_elements = best_solution_indices.shape[0]
best_solution_fitness = np.max(fitness)
#print("best_match_idx : ", best_match_idx)
#print("best_solution : ", best_solution)
#print("Selected indices : ", best_solution_indices)
print("Number of selected elements : ", best_solution_num_elements)
vector_output[:,:,fold] = get_vector(gen_labels,gen_preds)
population_output[:,:,fold] = harmony_memory
fold=fold+1
return population_output,vector_output