def violin_plot():
df = get_mod_data_original()
decoder = get_categorical_decoder_class()
df = df.replace({'CLASS': decoder})
df_columns = list(df.columns.values)
del df_columns[-1]
for col in df_columns:
ax = plt.figure(figsize=(14, 7))
ax = sns.violinplot(x='CLASS', y=df[col], data=df)
plt.show
def violin_plot_kdtree(df, y):
df['CLASS'] = y
decoder = get_categorical_decoder_class()
df = df.replace({'CLASS' : decoder})
df_columns = list(df.columns.values)
for col in df_columns:
ax = plt.figure(figsize=(14,7))
ax = sns.violinplot(x='CLASS', y=df[col], data=df)
img_name = str(col) + '.png'
plt.savefig(img_name, bbox_inches='tight')
def pca_2d(X, y, n_components=2):
pca = PCA(n_components=n_components)
pca_transform = pca.fit_transform(X)
X['pca_first'] = pca_transform[:, 0]
X['pca_second'] = pca_transform[:, 1]
X['CLASS'] = y
decoder = get_categorical_decoder_class()
X = X.replace({'CLASS': decoder})
print(pca.explained_variance_ratio_)
plt.figure(figsize=(15, 10))
sns.scatterplot(x="pca_first",
y="pca_second",
palette=sns.color_palette("hls", 7),
hue='CLASS',
data=X)
plt.show()
def tsne(X, y, perplexity, comp):
for perp in perplexity:
tsne = TSNE(n_components=comp, verbose=0, perplexity=perp, n_iter=250)
features = list(X.columns.values)
tsne_results = tsne.fit_transform(X[features].values)
X['tsne-2d-first'] = tsne_results[:, 0]
X['tsne-2d-second'] = tsne_results[:, 1]
X['CLASS'] = y
#X = X[X['CLASS'] != 0]
decoder = get_categorical_decoder_class()
X = X.replace({'CLASS': decoder})
print('Perplexity = ', perp)
plt.figure(figsize=(14, 10))
sns.scatterplot(x="tsne-2d-first",
y="tsne-2d-second",
hue='CLASS',
palette=sns.color_palette("hls", 7),
data=X,
legend="full",
alpha=0.3)
plt.show()
f1_avg += concensus[i]['f1']
# Predicciones
pos = len(X_estimar[0]) - 1
predictions_aux = model.predict(X_estimar[:, :pos].astype('float32'))
ids = get_estimar_ids()
for i in range(len(ids)):
if (ids[i] not in predictions):
predictions[ids[i]] = [int(predictions_aux[i])]
else:
predictions[ids[i]].append(int(predictions_aux[i]))
print('\nMÉTRICAS DEL MODELO (concenso)')
print('Accuracy: {}'.format(accuracy_avg / n))
print('Precision (macro): {}'.format(precision_avg / n))
print('Recall (macro): {}'.format(recall_avg / n))
print('F1 (macro): {}'.format(f1_avg / n))
# 6. Predicción final
# Diccionario para decodificar el nombre de las clases
categorical_decoder_class = get_categorical_decoder_class()
#Método que calcula la moda.
def most_frequent(lst):
return max(set(lst), key = lst.count)
with open(r'UPV_Astralaria.txt', 'w') as write_file:
write_file.write('ID|CLASE\n')
for sample in get_estimar_ids():
write_file.write('{}|{}\n'.format(sample, categorical_decoder_class[most_frequent(predictions[sample])]))
