[飞桨机器学习]SVM算法理论推导+代码实现
SVM之代码实现
首先我们回顾一下上一篇理论推导，并总结一下算法流程。
一、算法步骤
二、数据集
数据集依旧采用之前的TITANIC数据集。
三、代码实现
1.导入包
import csv
import numpy as np
import matplotlib.pyplot as plt
import copy
from time import sleep
import random
import types
2.数据处理
def loadDataset(filename):
with open(filename, 'r') as f:
lines = csv.reader(f)
data_set = list(lines)
if filename != 'titanic.csv':
for i in range(len(data_set)):
del(data_set[i][0])
# 整理数据
for i in range(len(data_set)):
del(data_set[i][0])
del(data_set[i][2])
data_set[i][4] += data_set[i][5]
del(data_set[i][5])
del(data_set[i][5])
del(data_set[i][6])
del(data_set[i][-1])
category = data_set[0]
del (data_set[0])
# 转换数据格式
for data in data_set:
data[0] = int(data[0])
data[1] = int(data[1])
if data[3] != '':
data[3] = float(data[3])
else:
data[3] = None
data[4] = float(data[4])
data[5] = float(data[5])
# 补全缺失值 转换记录方式 分类
for data in data_set:
if data[3] is None:
data[3] = 28
# male : 1, female : 0
if data[2] == 'male':
data[2] = 1
else:
data[2] = 0
# 经过测试，如果不将数据进行以下处理，分布会过于密集，处理后，数据的分布变得稀疏了
# age <25 为0, 25<=age<31为1，age>=31为2
if data[3] < 25:
data[3] = 0
elif data[3] >= 21 and data[3] < 60: # 但是测试得60分界准确率最高？？？！！！
data[3] = 1
else:
data[3] = 2
# sibsp&parcg以2为界限，小于为0，大于为1
if data[4] < 2:
data[4] = 0
else:
data[4] = 1
# fare以64为界限
if data[-1] < 64:
data[-1] = 0
else:
data[-1] = 1
return data_set, category
def split_data(data):
data_set = copy.deepcopy(data)
data_mat = []
label_mat = []
for i in range(len(data_set)):
if data_set[i][0] == 0:
data_set[i][0] = -1
label_mat.append(data_set[i][0])
del(data_set[i][0])
data_mat.append(data_set[i])
print(data_mat)
print(label_mat)
return data_mat, label_mat
3.SVM
def select_j_rand(i ,m):
# 选取alpha
j = i
while j == i:
j = int(random.uniform(0, m))
return j
def clip_alptha(aj, H, L):
# 修剪alpha
if aj > H:
aj = H
if L > aj:
aj = L
return aj
def smo(data_mat_In, class_label, C, toler, max_iter):
# 转化为numpy的mat存储
data_matrix = np.mat(data_mat_In)
label_mat = np.mat(class_label).transpose()
# data_matrix = data_mat_In
# label_mat = class_label
# 初始化b，统计data_matrix的纬度
b = 0
m, n = np.shape(data_matrix)
# 初始化alpha，设为0
alphas = np.mat(np.zeros((m, 1)))
# 初始化迭代次数
iter_num = 0
# 最多迭代max_iter次
while iter_num < max_iter:
alpha_pairs_changed = 0
for i in range(m):
# 计算误差Ei
fxi = float(np.multiply(alphas, label_mat).T*(data_matrix*data_matrix[i, :].T)) + b
Ei = fxi - float(label_mat[i])
# 优化alpha，松弛向量
if (label_mat[i]*Ei < -toler and alphas[i] < C) or (label_mat[i]*Ei > toler and alphas[i] > 0):
# 随机选取另一个与alpha_j成对优化的alpha_j
j = select_j_rand(i, m)
# 1.计算误差Ej
fxj = float(np.multiply(alphas, label_mat).T*(data_matrix*data_matrix[j, :].T)) + b
Ej = fxj - float(label_mat[j])
# 保存更新前的alpha，deepcopy
alpha_i_old = copy.deepcopy(alphas[i])
alpha_j_old = copy.deepcopy(alphas[j])
# 2.计算上下界L和H
if label_mat[i] != label_mat[j]:
L = max(0, alphas[j] - alphas[i])
H = min(C, C + alphas[j] - alphas[i])
else:
L = max(0, alphas[j] + alphas[i] - C)
H = min(C, alphas[j] + alphas[i])
if L == H:
print("L == H")
continue
# 3.计算eta
eta = 2.0 * data_matrix[i, :]*data_matrix[j, :].T - data_matrix[i, :]*data_matrix[i, :].T - data_matrix[j, :]*data_matrix[j, :].T
if eta >= 0:
print("eta >= 0")
continue
# 4.更新alpha_j
alphas[j] -= label_mat[j]*(Ei - Ej)/eta
# 5.修剪alpha_j
alphas[j] = clip_alptha(alphas[j], H, L)
if abs(alphas[j] - alphas[i]) < 0.001:
print("alpha_j变化太小")
continue
# 6.更新alpha_i
alphas[i] += label_mat[j]*label_mat[i]*(alpha_j_old - alphas[j])
# 7.更新b_1和b_2
b_1 = b - Ei - label_mat[i]*(alphas[i] - alpha_i_old)*data_matrix[i, :]*data_matrix[i, :].T - label_mat[j]*(alphas[j] - alpha_j_old)*data_matrix[i, :]*data_matrix[j, :].T
b_2 = b - Ej - label_mat[i]*(alphas[i] - alpha_i_old)*data_matrix[i, :]*data_matrix[j, :].T - label_mat[j]*(alphas[j] - alpha_j_old)*data_matrix[j, :] * data_matrix[j, :].T
# 8.根据b_1和b_2更新b
if 0 < alphas[i] and C > alphas[i]:
b = b_1
elif 0 < alphas[j] and C > alphas[j]:
b = b_2
else:
b = (b_1 + b_2)/2
# 统计优化次数
alpha_pairs_changed += 1
# 打印统计信息
print("第%d次迭代 样本：%d , alpha优化次数：%d" % (iter_num, i, alpha_pairs_changed))
# 更新迭代次数
if alpha_pairs_changed == 0:
iter_num += 1
else:
iter_num = 0
print("迭代次数：%d" % iter_num)
return b, alphas
def caluelate_w(data_mat, label_mat, alphas):
# 计算w
alphas = np.array(alphas)
data_mat = np.array(data_mat)
label_mat = np.array(label_mat)
# numpy.tile(A, reps):通过重复A给出的次数来构造数组。
# numpy中reshape函数的三种常见相关用法
# reshape(1, -1)转化成1行：
# reshape(2, -1)转换成两行：
# reshape(-1, 1)转换成1列：
# reshape(-1, 2)转化成两列
w = np.dot((np.tile(label_mat.reshape(1, -1).T, (1, 5))*data_mat).T, alphas)
return w.tolist()
4.预测
def prediction(test, w, b):
test = np.mat(test)
result = []
for i in test:
if i*w+b > 0:
result.append(1)
else:
result.append(-1)
print(result)
return result
5.主程序
if __name__ == "__main__":
test_set, category = loadDataset('titanic_test.csv')
data_set, category = loadDataset('titanic_train.csv')
test_mat, test_label = split_data(test_set)
data_mat, label_mat = split_data(data_set)
b, alphas = smo(data_mat, list(label_mat), 0.6, 0.001, 40)
print(b)
print(alphas)
w = caluelate_w(data_mat, label_mat, alphas)
print(w)
print(test_mat)
print(test_label)
result = prediction(test_mat, w, b)
count = 0
for i in range(len(result)):
if result[i] == test_label[i]:
count += 1
print(count)
print(count/len(result))
6.完整程序
import csv
import numpy as np
import matplotlib.pyplot as plt
import copy
from time import sleep
import random
import types
def loadDataset(filename):
with open(filename, 'r') as f:
lines = csv.reader(f)
data_set = list(lines)
if filename != 'titanic.csv':
for i in range(len(data_set)):
del(data_set[i][0])
# 整理数据
for i in range(len(data_set)):
del(data_set[i][0])
del(data_set[i][2])
data_set[i][4] += data_set[i][5]
del(data_set[i][5])
del(data_set[i][5])
del(data_set[i][6])
del(data_set[i][-1])
category = data_set[0]
del (data_set[0])
# 转换数据格式
for data in data_set:
data[0] = int(data[0])
data[1] = int(data[1])
if data[3] != '':
data[3] = float(data[3])
else:
data[3] = None
data[4] = float(data[4])
data[5] = float(data[5])
# 补全缺失值 转换记录方式 分类
for data in data_set:
if data[3] is None:
data[3] = 28
# male : 1, female : 0
if data[2] == 'male':
data[2] = 1
else:
data[2] = 0
# 经过测试，如果不将数据进行以下处理，分布会过于密集，处理后，数据的分布变得稀疏了
# age <25 为0, 25<=age<31为1，age>=31为2
if data[3] < 25:
data[3] = 0
elif data[3] >= 21 and data[3] < 60: # 但是测试得60分界准确率最高？？？！！！
data[3] = 1
else:
data[3] = 2
# sibsp&parcg以2为界限，小于为0，大于为1
if data[4] < 2:
data[4] = 0
else:
data[4] = 1
# fare以64为界限
if data[-1] < 64:
data[-1] = 0
else:
data[-1] = 1
return data_set, category
def split_data(data):
data_set = copy.deepcopy(data)
data_mat = []
label_mat = []
for i in range(len(data_set)):
if data_set[i][0] == 0:
data_set[i][0] = -1
label_mat.append(data_set[i][0])
del(data_set[i][0])
data_mat.append(data_set[i])
print(data_mat)
print(label_mat)
return data_mat, label_mat
def select_j_rand(i ,m):
# 选取alpha
j = i
while j == i:
j = int(random.uniform(0, m))
return j
def clip_alptha(aj, H, L):
# 修剪alpha
if aj > H:
aj = H
if L > aj:
aj = L
return aj
def smo(data_mat_In, class_label, C, toler, max_iter):
# 转化为numpy的mat存储
data_matrix = np.mat(data_mat_In)
label_mat = np.mat(class_label).transpose()
# data_matrix = data_mat_In
# label_mat = class_label
# 初始化b，统计data_matrix的纬度
b = 0
m, n = np.shape(data_matrix)
# 初始化alpha，设为0
alphas = np.mat(np.zeros((m, 1)))
# 初始化迭代次数
iter_num = 0
# 最多迭代max_iter次
while iter_num < max_iter:
alpha_pairs_changed = 0
for i in range(m):
# 计算误差Ei
fxi = float(np.multiply(alphas, label_mat).T*(data_matrix*data_matrix[i, :].T)) + b
Ei = fxi - float(label_mat[i])
# 优化alpha，松弛向量
if (label_mat[i]*Ei < -toler and alphas[i] < C) or (label_mat[i]*Ei > toler and alphas[i] > 0):
# 随机选取另一个与alpha_j成对优化的alpha_j
j = select_j_rand(i, m)
# 1.计算误差Ej
fxj = float(np.multiply(alphas, label_mat).T*(data_matrix*data_matrix[j, :].T)) + b
Ej = fxj - float(label_mat[j])
# 保存更新前的alpha，deepcopy
alpha_i_old = copy.deepcopy(alphas[i])
alpha_j_old = copy.deepcopy(alphas[j])
# 2.计算上下界L和H
if label_mat[i] != label_mat[j]:
L = max(0, alphas[j] - alphas[i])
H = min(C, C + alphas[j] - alphas[i])
else:
L = max(0, alphas[j] + alphas[i] - C)
H = min(C, alphas[j] + alphas[i])
if L == H:
print("L == H")
continue
# 3.计算eta
eta = 2.0 * data_matrix[i, :]*data_matrix[j, :].T - data_matrix[i, :]*data_matrix[i, :].T - data_matrix[j, :]*data_matrix[j, :].T
if eta >= 0:
print("eta >= 0")
continue
# 4.更新alpha_j
alphas[j] -= label_mat[j]*(Ei - Ej)/eta
# 5.修剪alpha_j
alphas[j] = clip_alptha(alphas[j], H, L)
if abs(alphas[j] - alphas[i]) < 0.001:
print("alpha_j变化太小")
continue
# 6.更新alpha_i
alphas[i] += label_mat[j]*label_mat[i]*(alpha_j_old - alphas[j])
# 7.更新b_1和b_2
b_1 = b - Ei - label_mat[i]*(alphas[i] - alpha_i_old)*data_matrix[i, :]*data_matrix[i, :].T - label_mat[j]*(alphas[j] - alpha_j_old)*data_matrix[i, :]*data_matrix[j, :].T
b_2 = b - Ej - label_mat[i]*(alphas[i] - alpha_i_old)*data_matrix[i, :]*data_matrix[j, :].T - label_mat[j]*(alphas[j] - alpha_j_old)*data_matrix[j, :] * data_matrix[j, :].T
# 8.根据b_1和b_2更新b
if 0 < alphas[i] and C > alphas[i]:
b = b_1
elif 0 < alphas[j] and C > alphas[j]:
b = b_2
else:
b = (b_1 + b_2)/2
# 统计优化次数
alpha_pairs_changed += 1
# 打印统计信息
print("第%d次迭代 样本：%d , alpha优化次数：%d" % (iter_num, i, alpha_pairs_changed))
# 更新迭代次数
if alpha_pairs_changed == 0:
iter_num += 1
else:
iter_num = 0
print("迭代次数：%d" % iter_num)
return b, alphas
def caluelate_w(data_mat, label_mat, alphas):
# 计算w
alphas = np.array(alphas)
data_mat = np.array(data_mat)
label_mat = np.array(label_mat)
# numpy.tile(A, reps):通过重复A给出的次数来构造数组。
# numpy中reshape函数的三种常见相关用法
# reshape(1, -1)转化成1行：
# reshape(2, -1)转换成两行：
# reshape(-1, 1)转换成1列：
# reshape(-1, 2)转化成两列
w = np.dot((np.tile(label_mat.reshape(1, -1).T, (1, 5))*data_mat).T, alphas)
return w.tolist()
def prediction(test, w, b):
test = np.mat(test)
result = []
for i in test:
if i*w+b > 0:
result.append(1)
else:
result.append(-1)
print(result)
return result
if __name__ == "__main__":
test_set, category = loadDataset('titanic_test.csv')
data_set, category = loadDataset('titanic_train.csv')
test_mat, test_label = split_data(test_set)
data_mat, label_mat = split_data(data_set)
b, alphas = smo(data_mat, list(label_mat), 0.6, 0.001, 40)
print(b)
print(alphas)
w = caluelate_w(data_mat, label_mat, alphas)
print(w)
print(test_mat)
print(test_label)
result = prediction(test_mat, w, b)
count = 0
for i in range(len(result)):
if result[i] == test_label[i]:
count += 1
print(count)
print(count/len(result))
三、测试结果
准确率：75.6%
本人正在学习人工智能方向，日常更新一些学习记录，欢迎互相交流
欢迎三连，互粉，等你呦
import csv
import numpy as np
import matplotlib.pyplot as plt
import copy
from time import sleep
import random
import types
def loadDataset(filename):
with open(filename, 'r') as f:
lines = csv.reader(f)
data_set = list(lines)
if filename != 'titanic.csv':
for i in range(len(data_set)):
del(data_set[i][0])
# 整理数据
for i in range(len(data_set)):
del(data_set[i][0])
del(data_set[i][2])
data_set[i][4] += data_set[i][5]
del(data_set[i][5])
del(data_set[i][5])
del(data_set[i][6])
del(data_set[i][-1])
category = data_set[0]
del (data_set[0])
# 转换数据格式
for data in data_set:
data[0] = int(data[0])
data[1] = int(data[1])
if data[3] != '':
data[3] = float(data[3])
else:
data[3] = None
data[4] = float(data[4])
data[5] = float(data[5])
# 补全缺失值 转换记录方式 分类
for data in data_set:
if data[3] is None:
data[3] = 28
# male : 1, female : 0
if data[2] == 'male':
data[2] = 1
else:
data[2] = 0
# 经过测试，如果不将数据进行以下处理，分布会过于密集，处理后，数据的分布变得稀疏了
# age <25 为0, 25<=age<31为1，age>=31为2
if data[3] < 25:
data[3] = 0
elif data[3] >= 21 and data[3] < 60: # 但是测试得60分界准确率最高？？？！！！
data[3] = 1
else:
data[3] = 2
# sibsp&parcg以2为界限，小于为0，大于为1
if data[4] < 2:
data[4] = 0
else:
data[4] = 1
# fare以64为界限
if data[-1] < 64:
data[-1] = 0
else:
data[-1] = 1
return data_set, category
def split_data(data):
data_set = copy.deepcopy(data)
data_mat = []
label_mat = []
for i in range(len(data_set)):
if data_set[i][0] == 0:
data_set[i][0] = -1
label_mat.append(data_set[i][0])
del(data_set[i][0])
data_mat.append(data_set[i])
print(data_mat)
print(label_mat)
return data_mat, label_mat
def select_j_rand(i ,m):
# 选取alpha
j = i
while j == i:
j = int(random.uniform(0, m))
return j
def clip_alptha(aj, H, L):
# 修剪alpha
if aj > H:
aj = H
if L > aj:
aj = L
return aj
def smo(data_mat_In, class_label, C, toler, max_iter):
# 转化为numpy的mat存储
data_matrix = np.mat(data_mat_In)
label_mat = np.mat(class_label).transpose()
# data_matrix = data_mat_In
# label_mat = class_label
# 初始化b，统计data_matrix的纬度
b = 0
m, n = np.shape(data_matrix)
# 初始化alpha，设为0
alphas = np.mat(np.zeros((m, 1)))
# 初始化迭代次数
iter_num = 0
# 最多迭代max_iter次
while iter_num < max_iter:
alpha_pairs_changed = 0
for i in range(m):
# 计算误差Ei
fxi = float(np.multiply(alphas, label_mat).T*(data_matrix*data_matrix[i, :].T)) + b
Ei = fxi - float(label_mat[i])
# 优化alpha，松弛向量
if (label_mat[i]*Ei < -toler and alphas[i] < C) or (label_mat[i]*Ei > toler and alphas[i] > 0):
# 随机选取另一个与alpha_j成对优化的alpha_j
j = select_j_rand(i, m)
# 1.计算误差Ej
fxj = float(np.multiply(alphas, label_mat).T*(data_matrix*data_matrix[j, :].T)) + b
Ej = fxj - float(label_mat[j])
# 保存更新前的alpha，deepcopy
alpha_i_old = copy.deepcopy(alphas[i])
alpha_j_old = copy.deepcopy(alphas[j])
# 2.计算上下界L和H
if label_mat[i] != label_mat[j]:
L = max(0, alphas[j] - alphas[i])
H = min(C, C + alphas[j] - alphas[i])
else:
L = max(0, alphas[j] + alphas[i] - C)
H = min(C, alphas[j] + alphas[i])
if L == H:
print("L == H")
continue
# 3.计算eta
eta = 2.0 * data_matrix[i, :]*data_matrix[j, :].T - data_matrix[i, :]*data_matrix[i, :].T - data_matrix[j, :]*data_matrix[j, :].T
if eta >= 0:
print("eta >= 0")
continue
# 4.更新alpha_j
alphas[j] -= label_mat[j]*(Ei - Ej)/eta
# 5.修剪alpha_j
alphas[j] = clip_alptha(alphas[j], H, L)
if abs(alphas[j] - alphas[i]) < 0.001:
print("alpha_j变化太小")
continue
# 6.更新alpha_i
alphas[i] += label_mat[j]*label_mat[i]*(alpha_j_old - alphas[j])
# 7.更新b_1和b_2
b_1 = b - Ei - label_mat[i]*(alphas[i] - alpha_i_old)*data_matrix[i, :]*data_matrix[i, :].T - label_mat[j]*(alphas[j] - alpha_j_old)*data_matrix[i, :]*data_matrix[j, :].T
b_2 = b - Ej - label_mat[i]*(alphas[i] - alpha_i_old)*data_matrix[i, :]*data_matrix[j, :].T - label_mat[j]*(alphas[j] - alpha_j_old)*data_matrix[j, :] * data_matrix[j, :].T
# 8.根据b_1和b_2更新b
if 0 < alphas[i] and C > alphas[i]:
b = b_1
elif 0 < alphas[j] and C > alphas[j]:
b = b_2
else:
b = (b_1 + b_2)/2
# 统计优化次数
alpha_pairs_changed += 1
# 打印统计信息
print("第%d次迭代 样本：%d , alpha优化次数：%d" % (iter_num, i, alpha_pairs_changed))
# 更新迭代次数
if alpha_pairs_changed == 0:
iter_num += 1
else:
iter_num = 0
print("迭代次数：%d" % iter_num)
return b, alphas
def caluelate_w(data_mat, label_mat, alphas):
# 计算w
alphas = np.array(alphas)
data_mat = np.array(data_mat)
label_mat = np.array(label_mat)
# numpy.tile(A, reps):通过重复A给出的次数来构造数组。
# numpy中reshape函数的三种常见相关用法
# reshape(1, -1)转化成1行：
# reshape(2, -1)转换成两行：
# reshape(-1, 1)转换成1列：
# reshape(-1, 2)转化成两列
w = np.dot((np.tile(label_mat.reshape(1, -1).T, (1, 5))*data_mat).T, alphas)
return w.tolist()
def prediction(test, w, b):
test = np.mat(test)
result = []
for i in test:
if i*w+b > 0:
result.append(1)
else:
result.append(-1)
print(result)
return result
if __name__ == "__main__":
test_set, category = loadDataset('data/data43004/titanic_test.csv')
data_set, category = loadDataset('data/data43004/titanic_train.csv')
test_mat, test_label = split_data(test_set)
data_mat, label_mat = split_data(data_set)
b, alphas = smo(data_mat, list(label_mat), 0.6, 0.001, 40)
print(b)
print(alphas)
w = caluelate_w(data_mat, label_mat, alphas)
print(w)
print(test_mat)
print(test_label)
result = prediction(test_mat, w, b)
count = 0
for i in range(len(result)):
if result[i] == test_label[i]:
count += 1
print(count)
print(count/len(result))
0.7606177606177607
运行代码请点击：https://aistudio.baidu.com/aistudio/projectdetail/619637?shared=1
欢迎三连！
1733
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
