-
Notifications
You must be signed in to change notification settings - Fork 3
/
svm_pso_predictor_mhTVAC.py
210 lines (179 loc) · 9.21 KB
/
svm_pso_predictor_mhTVAC.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
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
# coding: utf-8
import numpy as np
import random
import matplotlib.pyplot as plt
import pandas as pd
from sklearn import svm
from sklearn.model_selection import train_test_split
import cmath
from sklearn.model_selection import cross_val_score
from sklearn.model_selection import ShuffleSplit
# data = pd.read_csv('data/drug_cell/drug/AEW541_train_data-rfe.csv')
# X = data.iloc[:, :-1]
# y = data.iloc[:, -1]
# x_train, x_test, y_train, y_test = train_test_split(X, y, random_state=1, train_size=0.7)
#
# MAX_ITER = 500
# ----------------------PSO参数设置---------------------------------
class PSO_MHTVAC():
def __init__(self, max_iter, data_X, data_y, pN=30, dim=2, n_splits=3):
self.data_X = data_X
self.data_y = data_y
self.n_splits = n_splits
self.w = 0.9 # 惯性权重
self.w_start = 0.9
self.w_end = 0.4
self.c1 = 2
self.c2 = 2
self.c1f = 2.5
self.c1i = 0.5
self.c2f = 0.5
self.c2i = 2.5
self.r1 = random.uniform(0, 1)
self.r2 = random.uniform(0, 1)
self.r3 = random.uniform(0, 1)
self.r4 = random.uniform(0, 1)
self.r5 = random.uniform(0, 1)
# 突变相关的参数
self.mutation_num = 1 # 初始值暂定为种群的大小除以3
self.mutation_num_start = 1 # 初始值暂定为种群的大小除以3
self.mutation_num_end = 1 # 结束值暂定为种群的大小除以10
self.mprop = random.uniform(0, 1) # 突变概率
# self.mprop = 1 # 突变概率
self.rp = random.randint(0, pN - 1) # 随机选择一个微粒(index)
self.rd = random.randint(0, dim - 1) # 随机选择一个维度(index)
self.m = 2 # 常量, 怎么取值？？？
self.ms = 0.9 # 惯性权重
self.ms_start = 0.9
self.ms_end = 0.1
self.pN = pN # 粒子数量
self.dim = dim # 搜索维度
self.maxC = 1000 # 惩罚因子C的最大值
self.minC = 0.00001 # 惩罚因子C的最小值
self.maxGamma = 5 # 参数gamma的最大值
self.minGamma = 0.00001 # 参数gamma的最小值
self.max_v = np.array([self.maxC, self.maxGamma]) # 最大速度
self.min_v = np.array([-self.maxC, -self.maxGamma]) # 最小速度，反方向
self.max_x = np.array([self.maxC, self.maxGamma]) # 粒子位置的上界
self.min_x = np.array([self.minC, self.minGamma]) # 粒子位置的下界
# 随机速度
self.rv = random.uniform(0, self.max_v[random.randint(0, 1)])
self.max_iter = max_iter # 迭代次数
self.X = np.zeros((self.pN, self.dim)) # 所有粒子的位置和速度
self.V = np.zeros((self.pN, self.dim))
self.pbest = np.zeros((self.pN, self.dim)) # 个体经历的最佳位置和全局最佳位置
self.gbest = np.zeros((1, self.dim))
self.p_fit = np.zeros(self.pN) # 每个个体的历史最佳适应值
self.fit = 1e10 # 全局最佳适应值
# ---------------------目标函数Sphere函数-----------------------------
def function(self, c, g):
if g <= 0 or c <= 0:
return 1e10
model = svm.SVC(C=c, gamma=g) # gamma缺省值为 1.0/x.shape[1]
# model.fit(self.x_train, self.y_train)
# y_score = model.score(self.x_test, self.y_test)
cv = ShuffleSplit(n_splits=self.n_splits, test_size=.4, random_state=0)
score = cross_val_score(model, self.data_X, self.data_y, cv=cv)
print(score)
return -score.mean()
# ---------------------初始化种群----------------------------------
def init_Population(self):
for i in range(self.pN):
for j in range(self.dim):
self.X[i][j] = random.uniform(self.min_x[j], self.max_x[j]) # 位置的初始范围
self.V[i][j] = random.uniform(self.min_v[j], self.max_v[j]) # 速度的初始范围
self.pbest[i] = self.X[i]
tmp = self.function(self.X[i][0], self.X[i][1])
self.p_fit[i] = tmp
if tmp < self.fit:
self.fit = tmp
self.gbest = self.X[i]
# ----------------------更新粒子位置----------------------------------
def iterator(self):
fitness = []
for iter in range(self.max_iter):
for i in range(self.pN): # 更新gbest\pbest
temp = self.function(self.X[i][0], self.X[i][1])
if temp < self.p_fit[i]: # 更新个体最优
self.p_fit[i] = temp
self.pbest[i] = self.X[i]
if self.p_fit[i] < self.fit: # 更新全局最优
self.gbest = self.X[i]
self.fit = self.p_fit[i]
# 每次迭代都更新随机系数
self.r1 = random.uniform(0, 1)
self.r2 = random.uniform(0, 1)
self.r3 = random.uniform(0, 1)
self.r4 = random.uniform(0, 1)
self.r5 = random.uniform(0, 1)
for i in range(self.pN):
for d in range(self.dim): # 对维度遍历
# 去除惯性权重
self.V[i][d] = self.c1 * self.r1 * (self.pbest[i][d] - self.X[i][d]) + \
self.c2 * self.r2 * (self.gbest[d] - self.X[i][d])
self.rv = random.uniform(0, self.max_v[d])
if self.V[i][d] == 0:
if self.r3 < 0.5:
self.V[i][d] = self.r4 * self.rv
else:
self.V[i][d] = -self.r5 * self.rv
# 限制速度的简化代码
self.V[i][d] = np.sign(self.V[i][d]) * min(abs(self.V[i][d]), self.max_v[d])
self.X[i][d] = self.X[i][d] + self.V[i][d] # 更新粒子位置
# 限制粒子位置边界
if self.X[i][d] > self.max_x[d]:
self.X[i][d] = self.max_x[d]
elif self.X[i][d] < self.min_x[d]:
self.X[i][d] = self.min_x[d]
fitness.append(self.fit) # 追加全局最优
# 如果全局最优保持不变的话
if len(fitness) >= 2 and (fitness[-1] - fitness[-2] <= 0):
for mn in range(int(self.mutation_num)): # 突变的粒子数量
self.rp = random.randint(0, self.pN - 1) # 随机选择一个微粒(index)
self.rd = random.randint(0, self.dim - 1) # 随机选择一个维度(index)
if self.r1 < self.mprop:
if self.r2 < 0.5:
# self.V[self.rp][self.rd] += self.r3 * self.max_v[self.rd] / self.m
# 突变步长系数改为time varying, 大小用惯性权重
self.V[self.rp][self.rd] += self.ms * self.max_v[self.rd] / self.m
# 限制速度
self.V[self.rp][self.rd] = np.sign(self.V[self.rp][self.rd]) * min(
abs(self.V[self.rp][self.rd]), self.max_v[self.rd])
else:
# self.V[self.rp][self.rd] -= self.r4 * self.max_v[self.rd] / self.m
self.V[self.rp][self.rd] -= self.ms * self.max_v[self.rd] / self.m
# 限制速度
self.V[self.rp][self.rd] = np.sign(self.V[self.rp][self.rd]) * min(
abs(self.V[self.rp][self.rd]), self.max_v[self.rd])
print('V: %.5f,%.5f' % (self.V[0][0], self.V[0][1]), end="\t")
print('X: %.5f,%.5f' % (self.X[0][0], self.X[0][1]), end="\t")
print('fit: %.4f' % self.fit, end="\t") # 输出最优值
print('gBest: %.5f,%.5f' % (self.gbest[0], self.gbest[1]), end="\t") # 输出gBest
print('PSO-mhTVAC 当前迭代次数：', iter)
# 更新突变粒子个数
# self.mutation_num = self.mutation_num_start - \
# (self.mutation_num_start - self.mutation_num_end) * iter / self.max_iter
# 强转成整型
self.mutation_num = int(self.mutation_num)
# 更新学习因子
self.c1 = (self.c1f - self.c1i) * iter / self.max_iter + self.c1i
self.c2 = (self.c2f - self.c2i) * iter / self.max_iter + self.c2i
# 更新惯性权重
# self.w = self.w_start - (self.w_start - self.w_end) * iter / self.max_iter
# 更新突变步长系数
self.ms = self.ms_start - (self.ms_start - self.ms_end) * iter / self.max_iter
return fitness
# ----------------------程序执行-----------------------
# my_pso = PSO(pN=30, dim=2, max_iter=MAX_ITER) # 维度代表变量的个数
# my_pso.init_Population()
# fitness = my_pso.iterator()
# # -------------------画图--------------------
# plt.figure(1)
# plt.title("Figure1")
# plt.xlabel("iterators", size=14)
# plt.ylabel("fitness", size=14)
# t = np.array([t for t in range(0, MAX_ITER)])
# fitness = np.array(fitness)
# fitness_2 = [-v for v in fitness] # 取反，得到正数，模型准确率
# plt.plot(t, fitness_2, color='b', linewidth=3)
# plt.show()