from sklearn.datasets import make_blobs
from sklearn.cluster import KMeans

X, _ = make_blobs(n_samples=20, random_state=4)


def plot_KMeans(n):
    model = KMeans(n_clusters=2, init="random", n_init=1,
                   max_iter=n, random_state=6).fit(X)
    c0, c1 = model.cluster_centers_
    plt.scatter(X[model.labels_ == 0, 0], X[model.labels_ == 0, 1],
                marker='v', facecolor='r', edgecolors='k')
    plt.scatter(X[model.labels_ == 1, 0], X[model.labels_ == 1, 1],
                marker='^', facecolor='y', edgecolors='k')
    plt.scatter(c0[0], c0[1], marker='v', c="r", s=200)
    plt.scatter(c1[0], c1[1], marker='^', c="y", s=200)
    plt.grid(False)
    plt.title("반복횟수={}, 관성={:5.2f}".format(n, -model.score(X)))


plt.figure(figsize=(8, 8))
plt.subplot(321)
plot_KMeans(1)
plt.subplot(322)
plot_KMeans(2)
plt.subplot(323)
plot_KMeans(3)
plt.subplot(324)
plot_KMeans(4)
plt.tight_layout()
plt.show()

from sklearn.datasets import load_digits

digits = load_digits()
n_image = 20
np.random.seed(0)
idx = np.random.choice(range(len(digits.images)), n_image)
X = digits.data[idx]
images = digits.images[idx]

plt.figure(figsize=(12, 1))
for i in range(n_image):
    plt.subplot(1, n_image, i + 1)
    plt.imshow(images[i], cmap=plt.cm.bone)
    plt.grid(False)
    plt.xticks(())
    plt.yticks(())
    plt.title(i)

from scipy.cluster.hierarchy import linkage, dendrogram

Z = linkage(X, 'ward')
Z

from matplotlib.offsetbox import OffsetImage, AnnotationBbox

plt.figure(figsize=(10, 4))
ax = plt.subplot()

ddata = dendrogram(Z)

dcoord = np.array(ddata["dcoord"])
icoord = np.array(ddata["icoord"])
leaves = np.array(ddata["leaves"])
idx = np.argsort(dcoord[:, 2])
dcoord = dcoord[idx, :]
icoord = icoord[idx, :]
idx = np.argsort(Z[:, :2].ravel())
label_pos = icoord[:, 1:3].ravel()[idx][:20]

for i in range(20):
    imagebox = OffsetImage(images[i], cmap=plt.cm.bone_r, interpolation="bilinear", zoom=3)
    ab = AnnotationBbox(imagebox, (label_pos[i], 0),  box_alignment=(0.5, -0.1), 
                        bboxprops={"edgecolor" : "none"})
    ax.add_artist(ab)

plt.show()

from sklearn.datasets import make_circles, make_moons
from sklearn.cluster import DBSCAN

n_samples = 1000
np.random.seed(2)
X1, y1 = make_circles(n_samples=n_samples, factor=.5, noise=.09)
X2, y2 = make_moons(n_samples=n_samples, noise=.1)

def plot_DBSCAN(title, X, eps, xlim, ylim):
    model = DBSCAN(eps=eps)
    y_pred = model.fit_predict(X)
    idx_outlier = model.labels_ == -1
    plt.scatter(X[idx_outlier, 0], X[idx_outlier, 1], marker='x', lw=1, s=20)
    plt.scatter(X[model.labels_ == 0, 0], X[model.labels_ == 0, 1], marker='o', facecolor='g', s=5)
    plt.scatter(X[model.labels_ == 1, 0], X[model.labels_ == 1, 1], marker='s', facecolor='y', s=5)
    X_core = X[model.core_sample_indices_, :]
    idx_core_0 = np.array(list(set(np.where(model.labels_ == 0)[0]).intersection(model.core_sample_indices_)))
    idx_core_1 = np.array(list(set(np.where(model.labels_ == 1)[0]).intersection(model.core_sample_indices_)))
    plt.scatter(X[idx_core_0, 0], X[idx_core_0, 1], marker='o', facecolor='g', s=80, alpha=0.3)
    plt.scatter(X[idx_core_1, 0], X[idx_core_1, 1], marker='s', facecolor='y', s=80, alpha=0.3)
    plt.grid(False)
    plt.xlim(*xlim)
    plt.ylim(*ylim)
    plt.xticks(())
    plt.yticks(())
    plt.title(title)
    return y_pred

plt.figure(figsize=(10, 5))
plt.subplot(121)
y_pred1 = plot_DBSCAN("동심원 군집", X1, 0.1, (-1.2, 1.2), (-1.2, 1.2))
plt.subplot(122)
y_pred2 = plot_DBSCAN("초승달 군집", X2, 0.1, (-1.5, 2.5), (-0.8, 1.2))
plt.tight_layout()
plt.show()

from sklearn.metrics.cluster import adjusted_mutual_info_score, adjusted_rand_score

print("동심원 군집 ARI:", adjusted_rand_score(y1, y_pred1))
print("동심원 군집 AMI:", adjusted_mutual_info_score(y1, y_pred1))
print("초승달 군집 ARI:", adjusted_rand_score(y2, y_pred2))
print("초승달 군집 AMI:", adjusted_mutual_info_score(y2, y_pred2))