def fish_market():
# Read in Fish market
fish_data = pd.read_csv('data/fish-market/full.csv')
species = fish_data['Species'].value_counts().index.tolist()
# Test-train split
x, y = fish_data.drop('Species', axis=1).values, []
for fish in fish_data['Species']:
y.append(species.index(fish))
x_train, x_test, y_train, y_test = train_test_split(x,
y,
test_size=0.25,
random_state=1)
x_train, x_test = scale_data(x_train, x_test)
create_start = time.process_time()
# K-Means
k = 3
kmeans = KMeans(n_clusters=k).fit(x_train)
run_time = time.process_time() - create_start
print('Fish market [k-means] time (ms):', run_time * 1000)
km_as = accuracy_score(kmeans.predict(x_test), y_test)
print("k-means clustering accuracy score: ", km_as)
# Plot
plot_clusters(k, x_test, kmeans, 'clustering/k-means/fish-market')
def red_wine_quality():
# Read in Red Wine Quality
red_wine_data = pd.read_csv('data/red-wine-quality/full.csv')
# Label encode/transform
red_wine_data['quality'] = pd.cut(red_wine_data['quality'],
bins=[2, 5.5, 8],
labels=['bad', 'good'])
le = LabelEncoder()
red_wine_data['quality'] = le.fit_transform(red_wine_data['quality'])
# Test-train split
x = red_wine_data.drop('quality', axis=1).values
y = red_wine_data['quality']
x_train, x_test, y_train, y_test = train_test_split(x,
y,
test_size=0.2,
random_state=1)
x_train, x_test = scale_data(x_train, x_test)
create_start = time.process_time()
# K-Means
k = 2
kmeans = KMeans(n_clusters=k).fit(x_train)
run_time = time.process_time() - create_start
print('RWQ [k-means] time (ms):', run_time * 1000)
km_as = accuracy_score(kmeans.predict(x_test), y_test)
print("k-means clustering accuracy score: ", km_as)
# Plot
plot_clusters(k, x_test, kmeans, 'clustering/k-means/red-wine')
def fish_market():
# Read in Fish market
fish_data = pd.read_csv('data/fish-market/full.csv')
species = fish_data['Species'].value_counts().index.tolist()
# Test-train split
x, y = fish_data.drop('Species', axis=1).values, []
for fish in fish_data['Species']:
y.append(species.index(fish))
x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.25, random_state=1)
x_train, x_test = scale_data(x_train, x_test)
create_start = time.process_time()
# EM
k = 3
em = GaussianMixture(n_components=k).fit(x_train)
run_time = time.process_time() - create_start
print('Fish market [EM] time (ms):', run_time*1000)
em_as = accuracy_score(em.predict(x_test),y_test)
print("EM clustering accuracy score: ",em_as)
# Plot
plot_clusters(k, x_test, em, 'clustering/em/fish-market')
def plot_nn(x, y, suffix=''):
# Test-train split
x_train, x_test, y_train, y_test = train_test_split(x,
y,
test_size=0.2,
random_state=1)
x_train, x_test = scale_data(x_train, x_test)
iter_counts = np.arange(1, 1100, 50, dtype=int)
# Create Red Wine Quality NNs
nns = create_nns(x_train, y_train, iter_counts)
# Get training/testing accuracy for Red Wine Quality
accs_train, accs_test = [], []
for nn in nns:
accs_train.append(accuracy_score(nn.predict(x_train), y_train))
accs_test.append(accuracy_score(nn.predict(x_test), y_test))
# Generate graph for Red Wine Quality
plot = Plotter(name='red-wine{}'.format(suffix),
learner='nn',
axes={
'x': 'Number of weight updates',
'y': 'Accuracy score'
})
plot.add_plot(iter_counts, accs_train, 'training data', 'None')
plot.add_plot(iter_counts, accs_test, 'testing data', 'None')
plot.find_max(iter_counts, accs_test, 'testing')
plot.save()
def fish_market():
np.random.seed(93)
fish_data1, fish_data2 = pd.read_csv(
'data/fish-market/full.csv'), pd.read_csv('data/fish-market/full.csv')
x1, x2 = fish_data1.drop('Species', axis=1).values, fish_data2.drop(
['Species', 'Weight'], axis=1).values
y1 = fish_data1['Species']
# Label encode/transform
fish_data2['Weight'] = pd.cut(fish_data2['Weight'],
bins=[0, 120, 650, 1650],
labels=['light', 'avg', 'heavy'])
le = LabelEncoder()
fish_data2['Weight'] = le.fit_transform(fish_data2['Weight'].astype(str))
y2 = fish_data2['Weight']
x_train1, x_test1, y_train1, y_test1 = train_test_split(x1,
y1,
test_size=0.25,
random_state=1)
x_train2, x_test2, y_train2, y_test2 = train_test_split(x2,
y2,
test_size=0.25,
random_state=1)
x_train1, x_test1 = scale_data(x_train1, x_test1)
x_train2, x_test2 = scale_data(x_train2, x_test2)
k1 = KNeighborsClassifier(n_neighbors=5).fit(x_train1, y_train1)
k1_as = accuracy_score(k1.predict(x_test1), y_test1)
print("k1 accuracy score: ", k1_as)
k2 = KNeighborsClassifier(n_neighbors=5).fit(x_train2, y_train2)
k2_as = accuracy_score(k2.predict(x_test2), y_test2)
print("k2 accuracy score: ", k2_as)
def red_wine_quality(method='km'):
if method == 'km':
# Read in Red Wine Quality
red_wine_data_km = pd.read_csv('data/red-wine-quality/full.csv')
# Label encode/transform
red_wine_data_km['quality'] = pd.cut(red_wine_data_km['quality'],
bins=[2, 5.5, 8],
labels=['bad', 'good'])
le = LabelEncoder()
red_wine_data_km['quality'] = le.fit_transform(
red_wine_data_km['quality'])
# Group x, y
x = red_wine_data_km.drop('quality', axis=1).values
y = red_wine_data_km['quality']
create_start = time.process_time()
# PCA
c = 4
x = PCA(n_components=c).fit_transform(x)
test_train_start = time.process_time()
# Test-train split
x_train, x_test, y_train, y_test = train_test_split(x,
y,
test_size=0.2,
random_state=1)
x_train, x_test = scale_data(x_train, x_test)
test_train_time = time.process_time() - test_train_start
# K-Means
k = 2
kmeans = KMeans(n_clusters=k).fit(x_train)
run_time = time.process_time() - create_start - test_train_time
print('RWQ [PCA + k-means] time (ms):', run_time * 1000)
km_as = accuracy_score(kmeans.predict(x_test), y_test)
print("k-means clustering accuracy score: ", km_as)
# Plot
plot_clusters(k, x_test, kmeans,
'dimen-reduction/pca/red-wine-k-means')
elif method == 'em':
# Read in Red Wine Quality
red_wine_data_em = pd.read_csv('data/red-wine-quality/full.csv')
# Label encode/transform
red_wine_data_em['quality'] = pd.cut(red_wine_data_em['quality'],
bins=[2, 5.5, 8],
labels=['bad', 'good'])
le = LabelEncoder()
red_wine_data_em['quality'] = le.fit_transform(
red_wine_data_em['quality'])
# Group x, y
x = red_wine_data_em.drop('quality', axis=1).values
y = red_wine_data_em['quality']
create_start = time.process_time()
# PCA
c = 3
x = PCA(n_components=c).fit_transform(x)
test_train_start = time.process_time()
# Test-train split
x_train, x_test, y_train, y_test = train_test_split(x,
y,
test_size=0.2,
random_state=1)
x_train, x_test = scale_data(x_train, x_test)
test_train_time = time.process_time() - test_train_start
# EM
k = 2
em = GaussianMixture(n_components=k).fit(x_train)
run_time = time.process_time() - create_start - test_train_time
print('RWQ [PCA + EM] time (ms):', run_time * 1000)
em_as = accuracy_score(em.predict(x_test), y_test)
print("EM clustering accuracy score: ", em_as)
# Plot
plot_clusters(k, x_test, em, 'dimen-reduction/pca/red-wine-em')
else:
print('Invalid method: {}'.format(method))
def fish_market(method='km'):
if method == 'km':
# Read in Fish market
fish_data = pd.read_csv('data/fish-market/full.csv')
species = fish_data['Species'].value_counts().index.tolist()
# Group x, y
x, y = fish_data.drop('Species', axis=1).values, []
for fish in fish_data['Species']:
y.append(species.index(fish))
create_start = time.process_time()
# PCA
c = 6
x = PCA(n_components=c).fit_transform(x)
test_train_start = time.process_time()
# Test-train split
x_train, x_test, y_train, y_test = train_test_split(x,
y,
test_size=0.25,
random_state=1)
x_train, x_test = scale_data(x_train, x_test)
test_train_time = time.process_time() - test_train_start
# K-Means
k = 3
kmeans = KMeans(n_clusters=k).fit(x_train)
run_time = time.process_time() - create_start - test_train_time
print('Fish market [PCA + k-means] time (ms):', run_time * 1000)
km_as = accuracy_score(kmeans.predict(x_test), y_test)
print("k-means clustering accuracy score: ", km_as)
# Plot
plot_clusters(k, x_test, kmeans,
'dimen-reduction/pca/fish-market-k-means')
elif method == 'em':
# Read in Fish market
fish_data = pd.read_csv('data/fish-market/full.csv')
species = fish_data['Species'].value_counts().index.tolist()
# Group x, y
x, y = fish_data.drop('Species', axis=1).values, []
for fish in fish_data['Species']:
y.append(species.index(fish))
create_start = time.process_time()
# PCA
c = 3
x = PCA(n_components=c).fit_transform(x)
test_train_start = time.process_time()
# Test-train split
x_train, x_test, y_train, y_test = train_test_split(x,
y,
test_size=0.25,
random_state=1)
x_train, x_test = scale_data(x_train, x_test)
test_train_time = time.process_time() - test_train_start
# EM
k = 3
em = GaussianMixture(n_components=k).fit(x_train)
run_time = time.process_time() - create_start - test_train_time
print('Fish market [PCA + k-means] time (ms):', run_time * 1000)
em_as = accuracy_score(em.predict(x_test), y_test)
print("EM clustering accuracy score: ", em_as)
# Plot
plot_clusters(k, x_test, em, 'dimen-reduction/pca/fish-market-em')
else:
print('Invalid method: {}'.format(method))
