def part5_nn(clustering):
df = pd.DataFrame()
index = 0
history_dict = {}
# clusterings= ['Kmeans','EM']
alg_list = ['PCA', 'ICA', 'RP', 'RFE']
for alg in alg_list:
data_file = alg + os.sep + 'creditcard_{}_part5.csv'.format(clustering)
X, y = load_data.load_creditcard_data(data_file)
output_file = 'creditcard_{}_part5_{}_cm.png'.format(clustering, alg)
model, history, metrics, duration = nn(X,
y,
act='relu',
epoch=100,
output_file=output_file,
title=alg)
# print(model.metrics_names)
# print(metrics)
df.loc[index, 'alg'] = alg
df.loc[index, 'loss'] = metrics[0]
df.loc[index, 'acc'] = metrics[1]
df.loc[index, 'auc'] = metrics[2]
df.loc[index, 'duration'] = duration
index += 1
history_dict[alg] = history
df = df.round(3)
return df, history_dict
def part4():
df = pd.DataFrame()
index = 0
history_dict = {}
alg_list = ['PCA', 'ICA', 'RP', 'RFE', 'original']
for alg in alg_list:
if alg == 'original':
X, y = load_data.load_creditcard_data()
else:
data_file = alg + os.sep + 'creditcard.csv'
X, y = load_data.load_creditcard_data(data_file)
model, history, metrics, duration = nn(X, y, act='relu', epoch=100)
# print(model.metrics_names)
# print(metrics)
df.loc[index, 'alg'] = alg
df.loc[index, 'loss'] = metrics[0]
df.loc[index, 'acc'] = metrics[1]
df.loc[index, 'auc'] = metrics[2]
df.loc[index, 'duration'] = duration
index += 1
history_dict[alg] = history
df = df.round(3)
return df, history_dict
def kmeans(path, n_clusters=10):
if 'creditcard' in path:
X, y = load_data.load_creditcard_data(path)
else:
X, y = load_data.load_cancer_data(path)
model = KMeans(n_clusters=n_clusters, random_state=SEED).fit(X)
cluster_labels = model.fit_predict(X)
cluster_labels = cluster_labels.reshape(-1, 1)
enc = OneHotEncoder()
enc.fit(cluster_labels)
X_predict = enc.transform(cluster_labels)
X_predict = pd.DataFrame(X_predict.toarray())
output_file = path.split('.csv')[0] + '_Kmeans_part5.csv'
X_predict['label'] = y
X_predict.to_csv(output_file)
def rp(dataset, n_components, save_to_file=False, seed=1):
if dataset == 'creditcard':
X, y = load_data.load_creditcard_data()
else:
X, y = load_data.load_cancer_data()
model = GaussianRandomProjection(n_components=n_components,
random_state=seed)
model.fit(X)
## average log-likelihood of all samples
X_fitted = model.transform(X)
kurt = pd.DataFrame(X_fitted)
kurt = kurt.kurt(axis=0)
kurt = kurt.abs().mean()
reconstruction_error = util.reconstructionError(model, X)
if save_to_file:
X_fitted = pd.DataFrame(X_fitted)
X_fitted['label'] = y.values
X_fitted.to_csv(OUTPUT_DIR + os.sep + dataset + '.csv')
return kurt, reconstruction_error
def ica(dataset, n_components, save_to_file=False):
if dataset == 'creditcard':
X, y = load_data.load_creditcard_data()
else:
X, y = load_data.load_cancer_data()
model = FastICA(n_components=n_components, random_state=SEED).fit(X)
n_samples = X.shape[0]
## average log-likelihood of all samples
X_fitted = model.transform(X)
X_inverse = model.inverse_transform(X_fitted)
dist = np.linalg.norm(X - X_inverse) / n_samples
kurt = pd.DataFrame(X_fitted)
kurt = kurt.kurt(axis=0)
kurt = kurt.abs().mean()
if save_to_file:
X_fitted = pd.DataFrame(X_fitted)
X_fitted['label'] = y.values
X_fitted.to_csv(OUTPUT_DIR + os.sep + dataset + '.csv')
return dist, kurt
def GMM(path, n_clusters):
if 'creditcard' in path:
X, y = load_data.load_creditcard_data(path)
else:
X, y = load_data.load_cancer_data(path)
model = mixture.GaussianMixture(n_components=n_clusters,
random_state=SEED).fit(X)
cluster_labels = model.fit_predict(X)
silhouette_avg = silhouette_score(X, cluster_labels)
y_pred = model.predict(X)
cluster_acc = util.cluster_acc(y, y_pred)
return {
'silh': round(silhouette_avg, 3),
'cluster_acc': cluster_acc,
'aic': model.aic(X),
'bic': model.bic(X)
}
def pca(dataset, n_components, save_to_file=False):
if dataset == 'creditcard':
X, y = load_data.load_creditcard_data()
else:
X, y = load_data.load_cancer_data()
model = PCA(n_components=n_components, random_state=SEED).fit(X)
n_samples = X.shape[0]
## average log-likelihood of all samples
score = model.score(X)
variance = model.explained_variance_ratio_
cumsum = np.cumsum(model.explained_variance_ratio_)
X_fitted = model.transform(X)
X_inverse = model.inverse_transform(X_fitted)
dist = np.linalg.norm(X - X_inverse) / n_samples
if save_to_file:
X_fitted = pd.DataFrame(X_fitted)
X_fitted['label'] = y.values
X_fitted.to_csv(OUTPUT_DIR + os.sep + dataset + '.csv')
return score, variance, cumsum, dist
def kmeans(path, n_clusters):
if 'creditcard' in path:
X, y = load_data.load_creditcard_data(path)
else:
X, y = load_data.load_cancer_data(path)
kmeans = KMeans(n_clusters=n_clusters, random_state=SEED).fit(X)
n_samples = X.shape[0]
cluster_labels = kmeans.fit_predict(X)
silhouette_avg = silhouette_score(X, cluster_labels)
y_pred = kmeans.predict(X)
cluster_acc = util.cluster_acc(y, y_pred)
score = kmeans.score(X)
return {
'inertia': round(kmeans.inertia_ / n_samples, 1),
'silh': round(silhouette_avg, 3),
'cluster_acc': cluster_acc,
'score': score
}
def rfe(dataset, n_components, save_to_file=False):
if dataset == 'creditcard':
X, y = load_data.load_creditcard_data()
else:
X, y = load_data.load_cancer_data()
estimator = SVR(kernel="linear")
model = RFE(estimator, n_features_to_select=n_components, step=1)
model = model.fit(X, y)
n_samples = X.shape[0]
X_fitted = model.transform(X)
kurt = pd.DataFrame(X_fitted)
kurt = kurt.kurt(axis=0)
kurt = kurt.abs().mean()
X_inverse = model.inverse_transform(X_fitted)
reconstruction_error = np.linalg.norm(X - X_inverse) / n_samples
if save_to_file:
X_fitted = pd.DataFrame(X_fitted)
X_fitted['label'] = y.values
X_fitted.to_csv(OUTPUT_DIR + os.sep + dataset + '.csv')
return kurt, reconstruction_error
def gmm(dataset, n_components, n_init, max_iter, metric):
if dataset == 'creditcard':
X, y = load_data.load_creditcard_data()
else:
X, y = load_data.load_cancer_data()
model = mixture.GaussianMixture(n_components=n_components,
n_init=n_init,
max_iter=max_iter,
random_state=SEED).fit(X)
cluster_labels = model.fit_predict(X)
silhouette_avg = silhouette_score(X, cluster_labels, metric=metric)
y_pred = model.predict(X)
cluster_acc = util.cluster_acc(y, y_pred)
# return {'inertia':round(model.inertia_,1), 'silh':round(silhouette_avg,3)}
return {
'silh': round(silhouette_avg, 3),
'cluster_acc': cluster_acc,
'aic': model.aic(X),
'bic': model.bic(X)
}
def kmeans(dataset, n_clusters, n_init, max_iter, metric):
if dataset == 'creditcard':
X, y = load_data.load_creditcard_data()
else:
X, y = load_data.load_cancer_data()
kmeans = KMeans(n_clusters=n_clusters,
n_init=n_init,
max_iter=max_iter,
random_state=SEED).fit(X)
n_samples = X.shape[0]
cluster_labels = kmeans.fit_predict(X)
silhouette_avg = silhouette_score(X, cluster_labels, metric=metric)
y_pred = kmeans.predict(X)
cluster_acc = util.cluster_acc(y, y_pred)
score = kmeans.score(X)
return {
'inertia': round(kmeans.inertia_ / n_samples, 1),
'silh': round(silhouette_avg, 3),
'cluster_acc': cluster_acc,
'score': score
}