使用sklearn编写SVM程序

1. 环境

• python 3.7.2

• sklearn（python机器学习库）

  pip install scikit-learn
  

• numpy（矩阵处理库）

• matplotlib（绘图库）

2. 程序1

1. 程序1说明

   1. 程序主要功能为二维数据分成两类
   2. 数据类型（二维）：x坐标，y坐标，标签
   3. 输出：分类准确率，支持向量个数

2. 程序1代码

   # 用sklearn svm 训练rbf
   import os
   from sklearn.svm import SVC
   import matplotlib.pyplot as plt
   import numpy as np
   import csv
   from sklearn.model_selection import train_test_split, GridSearchCV
   path1 = os.path.abspath('.')
   
   
   # 加载数据
   def load(name):
       file = np.loadtxt(open(name, 'rb'), delimiter=',')
       x = file[:, [0, 1]]
       y = file[:, 2]
       return x, y
   
   
   if __name__ == "__main__":
       # 读取数据,针对二维线性不可分数据
       data, label = load('./data/train.csv')
       # 交叉验证4:1
       x_train, x_test, y_train, y_test = train_test_split(data, label, test_size=.2, random_state=0)
       # 设置参数
       c = 5000  # 惩罚参数
       g = 0.5  # 核函数参数
       # 初始化模型参数
       clf = SVC(cache_size=200, class_weight=None, coef0=0.0, C=c, decision_function_shape='ovr',
               degree=3, gamma=g, kernel='rbf', max_iter=-1, probability=False,
               random_state=None, shrinking=True, tol=0.001, verbose=False)
       clf.fit(x_train, y_train)
       # 预测data
       predict_list = clf.predict(x_test)
       # 预测精度
       precision = clf.score(x_test, y_test)
       print('第', i+1, '次 precision is : ', precision*100, "%")
       # 获取模型返回值
       n_Support_vector = clf.n_support_  # 支持向量个数
       print("支持向量个数为： ", n_Support_vector)
       Support_vector_index = clf.support_  # 支持向量索引
   
   

3. 程序2

1. 程序2在程序1的基础上添加了画图功能

2. 程序2代码

   # 用sklearn svm 训练，画图
   import os
   from sklearn.svm import SVC
   import matplotlib.pyplot as plt
   import numpy as np
   import csv
   from sklearn.model_selection import train_test_split
   path1 = os.path.abspath('.')
   
   
   # 加载数据
   def load(name):
       file = np.loadtxt(open(name, 'rb'), delimiter=',')
       x = file[:, [0, 1]]
       y = file[:, 2]
       return x, y
   
   
   # 用返回的参数绘制超平面
   def plot_point(data, label, Support_vector_index, clf, title):
       for i in range(np.shape(data)[0]):
           if label[i] == 1:
               plt.scatter(data[i][0], data[i][1], c='b', s=20)
           else:
               plt.scatter(data[i][0], data[i][1], c='y', s=20)
       for j in Support_vector_index:
           plt.scatter(data[j][0], data[j][1], s=100, c='', alpha=0.5, linewidth=1, edgecolor='g')
       # 画超平面
       x_min, x_max = data[:, 0].min() - 0.01, data[:, 0].max() + 0.01
       y_min, y_max = data[:, 1].min() - 0.01, data[:, 1].max() + 0.01
       h = 0.001
       xx, yy = np.meshgrid(np.arange(x_min, x_max, h), np.arange(y_min, y_max, h))
       plt.xlim(xx.min(), xx.max())
       plt.ylim(yy.min(), yy.max())
       plt.xticks(())
       plt.yticks(())
       z = clf.predict(np.c_[xx.ravel(), yy.ravel()])  # SVM的分割超平面
       # Put the result into a color plot
       z = z.reshape(xx.shape)
       plt.contourf(xx, yy, z, cmap='hot', alpha=0.3)
       plt.title(title)
       plt.show()
   
   
   if __name__ == "__main__":
       # 读取数据,针对二维线性不可分数据
       data, label = load('./data/train.csv')
       # 交叉验证
       x_train, x_test, y_train, y_test = train_test_split(data, label, test_size=.2, random_state=0)
       # 参数设置
       c = 1000
       g = 0.1
       # 初始化模型参数
       clf = SVC(cache_size=200, class_weight=None, coef0=0.0, C=c, decision_function_shape='ovr',
               degree=3, gamma=g, kernel='rbf', max_iter=-1, probability=False,
               random_state=None, shrinking=True, tol=0.001, verbose=False)
       clf.fit(x_train, y_train)
       # 预测data
       predict_list = clf.predict(x_test)
       # 预测精度
       precision = clf.score(x_test, y_test)
       print('第', i+1, '次 precision is : ', precision*100, "%")
       # 获取模型返回值
       n_Support_vector = clf.n_support_  # 支持向量个数
       print("支持向量个数为： ", n_Support_vector)
       Support_vector_index = clf.support_  # 支持向量索引
       title = 'C=' + str(c) + ', gamma=' + str(g)
       # 画图
       plot_point(data, label, Support_vector_index, clf, title)