SVM可视化分类平面、支持向量、bad case

# coding: utf-8 import numpy as np import csv from sklearn import svm import matplotlib.pyplot as plt from mpl_toolkits.mplot3d import Axes3D """ 采用 sklearn 中的svm ，以255占比值作为特征，对胸膜结节/肺内结节进行分类 """ train_path = './train_set.csv' test_path = './test_set.csv' def load_data(data_path): X,Y = [],[] csv_reader = csv.reader(open(data_path,'r')) for row in csv_reader: a = row[0][1:-1].split() X.append(np.array(a)) Y.append(np.array(row[1])) return X, Y def find_badcase(X, Y): bad_list = [] y = cls.predict(X) for i in range(len(X)): if y[i] != Y[i]: bad_list.append(i) return bad_list if __name__=="__main__": # load data x_train, y_train = load_data(train_path) x_test, y_test = load_data(test_path) # training cls = svm.SVC(kernel='linear', C=1.5) cls.fit(x_train, y_train) # accuracy print('Test score: %.4f' % cls.score(x_test, y_test)) print('Train score: %.4f' % cls.score(x_train, y_train)) # print bad case id bad_idx = find_badcase(x_test,y_test) n_Support_vector = cls.n_support_ # 支持向量个数 sv_idx = cls.support_ # 支持向量索引 w = cls.coef_ # 方向向量W b = cls.intercept_ # plot # 绘制分类平面 ax = plt.subplot(111, projection='3d') x = np.arange(0,1,0.01) y = np.arange(0,1,0.11) x, y = np.meshgrid(x, y) z = (w[0,0]*x + w[0,1]*y + b) / (-w[0,2]) surf = ax.plot_surface(x, y, z, rstride=1, cstride=1) # 绘制三维散点图 x_array = np.array(x_train, dtype=float) y_array = np.array(y_train, dtype=int) pos = x_array[np.where(y_array==1)] neg = x_array[np.where(y_array==-1)] ax.scatter(pos[:,0], pos[:,1], pos[:,2], c='r', label='pos') ax.scatter(neg[:,0], neg[:,1], neg[:,2], c='b', label='neg') # 绘制支持向量 X = np.array(x_train,dtype=float) for i in range(len(sv_idx)): ax.scatter(X[sv_idx[i],0], X[sv_idx[i],1], X[sv_idx[i],2],s=50, c='',marker='o', edgecolors='g') # 绘制 bad case # x_test = np.array(x_test,dtype=float) # for i in range(len(bad_idx)): # j = bad_idx[i] # ax.scatter(x_test[j,0], x_test[j,1], x_test[j,2],s=60, # c='',marker='o', edgecolors='g') ax.set_zlabel('Z') # 坐标轴 ax.set_ylabel('Y') ax.set_xlabel('X') ax.set_zlim([0, 1]) plt.legend(loc='upper left') ax.view_init(35,300) plt.show()