非監督式學習

 1: # 隨機生成100個(x, y)
 2: import pandas as pd
 3: import matplotlib.pyplot as plt
 4: 
 5: data = {
 6:     'x': [25, 34, 22, 27, 33, 33, 31, 22, 35, 34, 67, 54, 57, 43, 50, 57, 59, 52, 65, 47, 49, 48, 35, 33, 44, 45, 38,
 7:           43, 51, 46],
 8:     'y': [79, 51, 53, 78, 59, 74, 73, 57, 69, 75, 51, 32, 40, 47, 53, 36, 35, 58, 59, 50, 25, 20, 14, 12, 20, 5, 29, 27,
 9:           8, 7]
10:     }
11: samples = pd.DataFrame(data)
12: plt.scatter(samples['x'], samples['y'])
13: plt.savefig("images/kmeansScatterData.png", dpi=300)
14: #plt.show()

 1: # 隨機生成100個(x, y)
 2: import pandas as pd
 3: data = {
 4:     'x': [25, 34, 22, 27, 33, 33, 31, 22, 35, 34, 67, 54, 57, 43, 50, 57, 59, 52, 65, 47, 49, 48, 35, 33, 44, 45, 38,
 5:           43, 51, 46],
 6:     'y': [79, 51, 53, 78, 59, 74, 73, 57, 69, 75, 51, 32, 40, 47, 53, 36, 35, 58, 59, 50, 25, 20, 14, 12, 20, 5, 29, 27,
 7:           8, 7]
 8:     }
 9: samples = pd.DataFrame(data)
10: 
11: import matplotlib.pyplot as plt
12: from sklearn.cluster import KMeans
13: 
14: kmeans = KMeans(n_clusters=3) #預計分為三群，迭代次數由模型自行定義
15: kmeans.fit(samples)
16: cluster = kmeans.predict(samples)
17: 
18: plt.scatter(samples['x'], samples['y'], c=cluster, cmap=plt.cm.Set1)
19: plt.savefig("images/kmeansScatter.png", dpi=300)
20: #plt.show()

 1: #輸出說明
 2: import numpy as np
 3: import matplotlib.pyplot as plt
 4: from matplotlib import image
 5: from sklearn.cluster import MiniBatchKMeans
 6: 
 7: # K 值 (要保留的顏色數量)
 8: K = 4
 9: 
10: # 讀取圖片
11: image = image.imread(r'./images/Photo42.jpg')
12: w, h, d = tuple(image.shape)
13: print(w,h,d)
14: image_data = np.reshape(image, (w * h, d))/ 255
15: print(image_data.shape)
16: print(image_data[0])
17: # 將顏色分類為 K 種
18: kmeans = MiniBatchKMeans(n_clusters=K, batch_size=10)
19: labels = kmeans.fit_predict(image_data)
20: print(labels[:10])
21: centers = kmeans.cluster_centers_
22: print(centers.shape)
23: print(centers) #K個中心點的RGB
24: # 根據分類將顏色寫入新的影像陣列
25: image_compressed = np.zeros(image.shape)
26: label_idx = 0
27: for i in range(w):
28:   for j in range(h):
29:     image_compressed[i][j] = centers[labels[label_idx]]
30:     label_idx += 1
31: 
32: plt.imsave(r'images/compressTest.jpg', image_compressed)

 1:   import pandas as pd
 2:   from sklearn.model_selection import train_test_split
 3:   from sklearn.preprocessing import StandardScaler
 4:   import numpy as np
 5:   import matplotlib.pyplot as plt
 6:   from sklearn.decomposition import PCA
 7: 
 8:   # ## Extracting the principal components step-by-step
 9: 
10:   df_wine = pd.read_csv('https://archive.ics.uci.edu/ml/'
11:                         'machine-learning-databases/wine/wine.data',
12:                         header=None)
13: 
14:   df_wine.columns = ['Class label', 'Alcohol', 'Malic acid', 'Ash',
15:                      'Alcalinity of ash', 'Magnesium', 'Total phenols',
16:                      'Flavanoids', 'Nonflavanoid phenols', 'Proanthocyanins',
17:                      'Color intensity', 'Hue',
18:                      'OD280/OD315 of diluted wines', 'Proline']
19: 
20:   print(df_wine.head())
21: 
22:   # Splitting the data into 70% training and 30% test subsets.
23: 
24:   X, y = df_wine.iloc[:, 1:].values, df_wine.iloc[:, 0].values
25: 
26:   X_train, X_test, y_train, y_test = train_test_split(X, y,
27:                                      test_size=0.3,
28:                                      stratify=y, random_state=0)
29: 
30:   # 1. Standardizing the data.
31:   sc = StandardScaler()
32:   X_train_std = sc.fit_transform(X_train)
33:   X_test_std = sc.transform(X_test)
34: 
35:   # 2. Eigendecomposition of the covariance matrix.
36:   cov_mat = np.cov(X_train_std.T)
37:   eigen_vals, eigen_vecs = np.linalg.eig(cov_mat)
38: 
39:   print('\nEigenvalues \n%s' % eigen_vals)
40: 
41:   # ## Total and explained variance
42: 
43:   tot = sum(eigen_vals)
44:   var_exp = [(i / tot) for i in sorted(eigen_vals, reverse=True)]
45:   cum_var_exp = np.cumsum(var_exp)
46: 
47:   plt.bar(range(1, 14), var_exp, alpha=0.5, align='center',
48:           label='individual explained variance')
49:   plt.step(range(1, 14), cum_var_exp, where='mid',
50:            label='cumulative explained variance')
51:   plt.ylabel('Explained variance ratio')
52:   plt.xlabel('Principal component index')
53:   plt.legend(loc='best')
54:   plt.tight_layout()
55:   plt.savefig('05_02.png', dpi=300)
56:   #plt.show()
57: 

 1:   import pandas as pd
 2:   from sklearn.model_selection import train_test_split
 3:   from sklearn.preprocessing import StandardScaler
 4:   import numpy as np
 5:   import matplotlib.pyplot as plt
 6:   from sklearn.decomposition import PCA
 7: 
 8:   # ## Extracting the principal components step-by-step
 9: 
10:   df_wine = pd.read_csv('https://archive.ics.uci.edu/ml/'
11:                           'machine-learning-databases/wine/wine.data',
12:                           header=None)
13: 
14:   #  df_wine.columns = ['Class label', 'Alcohol', 'Malic acid', 'Ash',
15:   #                     'Alcalinity of ash', 'Magnesium', 'Total phenols',
16:   #                     'Flavanoids', 'Nonflavanoid phenols', 'Proanthocyanins',
17:   #                     'Color intensity', 'Hue',
18:   #                     'OD280/OD315 of diluted wines', 'Proline']
19: 
20:       # Splitting the data into 70% training and 30% test subsets.
21:   X, y = df_wine.iloc[:, 1:].values, df_wine.iloc[:, 0].values
22:   X_train, X_test, y_train, y_test = train_test_split(X, y,
23:                                       test_size=0.3,
24:                                       stratify=y, random_state=0)
25:   # 1. Standardizing the data.
26:   sc = StandardScaler()
27:   X_train_std = sc.fit_transform(X_train)
28:   X_test_std = sc.transform(X_test)
29:   # 2. Eigendecomposition of the covariance matrix.
30:   cov_mat = np.cov(X_train_std.T)
31:   eigen_vals, eigen_vecs = np.linalg.eig(cov_mat)
32:   # ## Total and explained variance
33:   #tot = sum(eigen_vals)
34:   #var_exp = [(i / tot) for i in sorted(eigen_vals, reverse=True)]
35:   #cum_var_exp = np.cumsum(var_exp)
36:   # ## Feature transformation
37:   # Make a list of (eigenvalue, eigenvector) tuples
38:   eigen_pairs = [(np.abs(eigen_vals[i]), eigen_vecs[:, i])
39:                   for i in range(len(eigen_vals))]
40:   # Sort the (eigenvalue, eigenvector) tuples from high to low
41:   eigen_pairs.sort(key=lambda k: k[0], reverse=True)
42:   w = np.hstack((eigen_pairs[0][1][:, np.newaxis],
43:                   eigen_pairs[1][1][:, np.newaxis]))
44:   print('Matrix W:\n', w)
45:   print(X_train_std[0].dot(w))
46:   X_train_pca = X_train_std.dot(w)
47:   # plot
48:   colors = ['r', 'b', 'g']
49:   markers = ['s', 'x', 'o']
50: 
51:   for l, c, m in zip(np.unique(y_train), colors, markers):
52:       plt.scatter(X_train_pca[y_train == l, 0],
53:                   X_train_pca[y_train == l, 1],
54:                   c=c, label=l, marker=m)
55: 
56:   plt.xlabel('PC 1')
57:   plt.ylabel('PC 2')
58:   plt.legend(loc='lower left')
59:   plt.tight_layout()
60:   plt.savefig('05_03.png', dpi=300)
61:   #plt.show()
62: 

 1:   from matplotlib.colors import ListedColormap
 2:   import pandas as pd
 3:   from sklearn.model_selection import train_test_split
 4:   from sklearn.preprocessing import StandardScaler
 5:   import numpy as np
 6:   import matplotlib.pyplot as plt
 7:   from sklearn.decomposition import PCA
 8:   from sklearn.linear_model import LogisticRegression
 9: 
10:   # ## Extracting the principal components step-by-step
11: 
12:   df_wine = pd.read_csv('https://archive.ics.uci.edu/ml/'
13:                           'machine-learning-databases/wine/wine.data',
14:                           header=None)
15: 
16:   #  df_wine.columns = ['Class label', 'Alcohol', 'Malic acid', 'Ash',
17:   #                     'Alcalinity of ash', 'Magnesium', 'Total phenols',
18:   #                     'Flavanoids', 'Nonflavanoid phenols', 'Proanthocyanins',
19:   #                     'Color intensity', 'Hue',
20:   #                     'OD280/OD315 of diluted wines', 'Proline']
21: 
22:       # Splitting the data into 70% training and 30% test subsets.
23:   X, y = df_wine.iloc[:, 1:].values, df_wine.iloc[:, 0].values
24:   X_train, X_test, y_train, y_test = train_test_split(X, y,
25:                                       test_size=0.3,
26:                                       stratify=y, random_state=0)
27:   # 1. Standardizing the data.
28:   sc = StandardScaler()
29:   X_train_std = sc.fit_transform(X_train)
30:   X_test_std = sc.transform(X_test)
31: 
32:   def plot_decision_regions(X, y, classifier, resolution=0.02):
33:       # setup marker generator and color map
34:       markers = ('s', 'x', 'o', '^', 'v')
35:       colors = ('red', 'blue', 'lightgreen', 'gray', 'cyan')
36:       cmap = ListedColormap(colors[:len(np.unique(y))])
37: 
38:       # plot the decision surface
39:       x1_min, x1_max = X[:, 0].min() - 1, X[:, 0].max() + 1
40:       x2_min, x2_max = X[:, 1].min() - 1, X[:, 1].max() + 1
41:       xx1, xx2 = np.meshgrid(np.arange(x1_min, x1_max, resolution),
42:                              np.arange(x2_min, x2_max, resolution))
43:       Z = classifier.predict(np.array([xx1.ravel(), xx2.ravel()]).T)
44:       Z = Z.reshape(xx1.shape)
45:       plt.contourf(xx1, xx2, Z, alpha=0.4, cmap=cmap)
46:       plt.xlim(xx1.min(), xx1.max())
47:       plt.ylim(xx2.min(), xx2.max())
48: 
49:       # plot class samples
50:       for idx, cl in enumerate(np.unique(y)):
51:           plt.scatter(x=X[y == cl, 0],
52:                       y=X[y == cl, 1],
53:                       alpha=0.6,
54:                       c=cmap(idx),
55:                       edgecolor='black',
56:                       marker=markers[idx],
57:                       label=cl)
58: 
59:   # Training logistic regression classifier using the first 2 principal components.
60:   pca = PCA(n_components=2)
61:   X_train_pca = pca.fit_transform(X_train_std)
62:   X_test_pca = pca.transform(X_test_std)
63: 
64:   lr = LogisticRegression()
65:   lr = lr.fit(X_train_pca, y_train)
66: 
67:   plot_decision_regions(X_train_pca, y_train, classifier=lr)
68:   plt.xlabel('PC 1')
69:   plt.ylabel('PC 2')
70:   plt.legend(loc='lower left')
71:   plt.tight_layout()
72:   plt.savefig('05_04.png', dpi=300)
73:   #plt.show()
74:   plot_decision_regions(X_test_pca, y_test, classifier=lr)
75:   plt.xlabel('PC 1')
76:   plt.ylabel('PC 2')
77:   plt.legend(loc='lower left')
78:   plt.tight_layout()
79:   plt.savefig('05_05.png', dpi=300)
80:   #plt.show()