def exercicio7():
utils.print_header(7)
np.random.seed(constants.SEED)
print('Spiral')
spiral = load_database(constants.FILENAME_SPIRAL_BATABASE,
constants.URL_SPIRAL_BATABASE)
print('\t- Nb samples: {}'.format(spiral.shape[0]))
print('\n\tkMeans:')
kmeans_clusters = k_means(spiral, 3, max_iters=300, db_name='spiral')
print('\t\t- Purity (kMeans): {:.2f}%'.format(
utils.purity(kmeans_clusters)))
print('\t\t- Dist_intra_inter (kMeans): {:.2f}%'.format(
utils.dist_intra_inter(kmeans_clusters)))
print('\n\tAGNES:')
agnes_clusters = agnes(spiral, 3, db_name='spiral')
print('\t\t- Purity (AGNES): {:.2f}%'.format(utils.purity(agnes_clusters)))
print('\t\t- Dist_intra_inter (AGNES): {:.2f}%'.format(
utils.dist_intra_inter(agnes_clusters)))
print('\nJain')
jain = load_database(constants.FILENAME_JAIN_DATABASE,
constants.URL_JAIN_DATABASE)
print('\t- Nb samples: {}'.format(jain.shape[0]))
print('\n\t- kMeans:')
clusters = k_means(jain, 2, max_iters=300, db_name='jain')
print('\t\t- Purity (kMeans): {:.2f}%'.format(utils.purity(clusters)))
print('\t\t- Dist_intra_inter (kMeans): {:.2f}%'.format(
utils.dist_intra_inter(clusters)))
print('\n\tAGNES:')
agnes_clusters = agnes(jain, 2, db_name='jain')
print('\t\t- Purity (AGNES): {:.2f}%'.format(utils.purity(agnes_clusters)))
print('\t\t- Dist_intra_inter (AGNES): {:.2f}%'.format(
utils.dist_intra_inter(agnes_clusters)))
exit()
# feature extraction
digits = utils.resize(digits, 16)
digits = utils.get_deskew_imgs(digits)
# pix = utils.get_pix_features(digits)
hog = utils.get_hog_feature(digits)
# X = np.hstack([pix])
X = np.hstack([hog])
X = utils.normalize(X)
#-------------
# data reduction
pca = pca.PCA(X)
n_component = 0
gained = 0
while (gained < pov):
gained += pca.pov(n_component)
n_component += 1
X_reduced = pca.transform(X, n_component)
#-------------
# clustring
start_time = timeit.default_timer()
print("clustring with k-means with k =", classes, "...")
k_means = kmeans.kmeans(X_reduced, y)
clusters, centers = k_means.train(classes, iter)
print("K-means purity:", utils.purity(clusters, y, classes))
print("K-means rand-index:", utils.rand_index(clusters, y, classes))
elapsed = timeit.default_timer() - start_time
print("execution time: ",
str(int((elapsed / 60))) + ':' + str(int(elapsed % 60)))
import cProfile
import numpy as np
from rock import RockClustering
from utils import categorical_to_binary, purity
if __name__ == '__main__':
data = np.loadtxt("data/agaricus-lepiota.data", dtype=str, delimiter=",", skiprows=0)
for i, neighborhood in enumerate(np.unique(data[:, 0]), 0):
data[data[:, 0] == neighborhood, 0] = i
labels = np.asarray(data[:, 0], dtype=int)
data = np.asarray(data[:, 1:])
profile = cProfile.Profile()
profile.enable()
clustering = RockClustering(categorical_to_binary(data[:, :]), 20, theta=0.80)
final_clusters = clustering.clusters()
profile.disable()
profile.print_stats(sort='cumulative')
for i, cluster in enumerate(final_clusters, 1):
print("Cluster no. {},\nlength = {}".format(i, len(cluster.points)))
print(labels[cluster.points])
print("Purity = {}".format(purity(final_clusters, labels[:])))
import cProfile
import numpy as np
from rock import RockRealClustering
from utils import categorical_to_binary, purity
from sklearn.preprocessing import MinMaxScaler
if __name__ == '__main__':
data = np.loadtxt("data/iris.data", dtype=str, delimiter=",", skiprows=0)
labels = np.asarray(data[:, -1], dtype=str)
integer_labels = np.zeros(labels.shape[0], dtype=int)
for i, label in enumerate(np.unique(labels), 0):
integer_labels[labels == label] = i
data = np.asarray(data[:, :-1], dtype=float)
data = MinMaxScaler().fit_transform(data)
profile = cProfile.Profile()
profile.enable()
clustering = RockRealClustering(data, 3, theta=0.50, nbr_max_distance=0.20)
final_clusters = clustering.clusters()
profile.disable()
profile.print_stats(sort='time')
for i, cluster in enumerate(final_clusters, 1):
print("Cluster no. {},\nlength = {}".format(i, len(cluster.points)))
print(labels[cluster.points])
print("Purity = {}".format(purity(final_clusters, integer_labels)))
profile = cProfile.Profile()
profile.enable()
clustering = RockGeoClustering(data, 5, theta=0.5, nbr_max_distance=40)
final_clusters = clustering.clusters()
profile.disable()
profile.print_stats(sort='time')
for i, cluster in enumerate(final_clusters, 1):
print("Cluster no. {},\nlength = {}".format(i, len(cluster.points)))
print(labels[cluster.points])
counts = np.bincount(np.asarray(labels, dtype=int)[cluster.points])
dominant = np.argmax(counts)
if len(cluster.points) != 1:
cluster_color = colors[random.randint(0, len(colors) - 1)]
for point in cluster.points:
folium.Marker(location=data[point, :],
icon=folium.Icon(color=cluster_color,
icon='circle',
prefix='fa-')).add_to(map)
else:
for point in cluster.points:
folium.Marker(location=data[point, :],
icon=folium.Icon(color="purple",
icon='circle',
prefix='fa-')).add_to(map)
print("Purity = {}".format(
purity(final_clusters, np.asarray(labels, dtype=int))))
map.save('docs/nyc_clustering.html')