def cross_val_group(data, msk, folds=10):
train, test = split_for_train_test(data, msk, 2)
fold_size = 0
if len(train) > 1:
fold_size = floor(len(train[0]) / folds)
else:
fold_size = floor(len(train) / folds)
# msk = rand(fold_size) < acc
pool = range(folds)
list_folds = []
for i in range(fold_size):
l_fold = sample(list(pool), k=len(pool))
for j in range(len(l_fold)):
list_folds.append(l_fold[j])
return train, test, list_folds
# print(row[13])
if row[2] == '1':
spiral_1[0].append(float(row[0]))
spiral_1[1].append(float(row[1]))
spiral_t.append(row[2])
elif row[2] == '2':
spiral_2[0].append(float(row[0]))
spiral_2[1].append(float(row[1]))
spiral_t.append(row[2])
else:
print(row)
folds = 10
msk = rand(len(spiral_1[0])) < 0.9
sp1_train, sp1_test, sp_list = cross_val_group(spiral_1, msk)
sp2_train, sp2_test = split_for_train_test(spiral_2, msk, 2)
spt_train, spt_test = split_for_train_test(spiral_t, msk)
# print(spt_test)
cross_rslts = []
folds_dict = {}
for i in range(folds):
rslts = []
train_test_fold = array([x != i for x in sp_list])
train_fold_1 = [
list(compress(sp1_train[0], train_test_fold)),
list(compress(sp1_train[1], train_test_fold))
]
test_fold_1 = [
list(compress(sp1_train[0], ~train_test_fold)),
'C:\\Users\\vinic\\OneDrive\\Faculdade\\Reconhecimento de padrões\\exercicios\\heart.csv',
'r',
newline='') as csv_heart:
csv_reader = csv.reader(csv_heart, delimiter=',')
for row in csv_reader:
# print(row[13])
if row[13] == '1':
heart_1.append([float(row[i]) for i in range(len(row) - 1)])
elif row[13] == '2':
heart_2.append([float(row[i]) for i in range(len(row) - 1)])
else:
print(row)
msk = rand(len(heart_1)) < tax
train_1, test_1 = split_for_train_test(heart_1, msk)
train_2, test_2 = split_for_train_test(heart_2, msk)
priori_1 = len(heart_1) / (len(heart_1) + len(heart_2))
priori_2 = len(heart_2) / (len(heart_1) + len(heart_2))
u1 = mean(train_1, axis=0)
u2 = mean(train_2, axis=0)
cov1 = cov(train_1, rowvar=False)
cov2 = cov(train_2, rowvar=False)
rslt = []
rslt.append(
classifier_N(test_1,
u1,
dist_y = normal(4, std_d_2, points)
dist_p2 = [dist_x, dist_y]
dist_x = normal(2, std_d_3, points)
dist_y = normal(4, std_d_3, points)
dist_p3 = [dist_x, dist_y]
dist_x = normal(4, std_d_4, points)
dist_y = normal(2, std_d_4, points)
dist_p4 = [dist_x, dist_y]
# separação dos dados
msk = rand(points) < 0.9
train_1, test_1 = split_for_train_test(dist_p1, msk, 2)
train_2, test_2 = split_for_train_test(dist_p2, msk, 2)
train_3, test_3 = split_for_train_test(dist_p3, msk, 2)
train_4, test_4 = split_for_train_test(dist_p4, msk, 2)
# treinamento
u = []
s = []
u.append(mean(train_1, axis=1))
u.append(mean(train_2, axis=1))
u.append(mean(train_3, axis=1))
u.append(mean(train_4, axis=1))
s.append(std(train_1, axis=1))
s.append(std(train_2, axis=1))
concatenate((dist_2_x, dist_3_x)),
concatenate((dist_2_y, dist_3_y))
]
dist_4_x = normal(-2, std_d_2, points)
dist_4_y = normal(-2, std_d_2, points)
dist_1_x = normal(2, std_d_2, points)
dist_1_y = normal(2, std_d_2, points)
dist_p2 = [
concatenate((dist_1_x, dist_4_x)),
concatenate((dist_1_y, dist_4_y))
]
msk = rand(points * 2) < 0.9
train_x, test_x = split_for_train_test(dist_p1[0], msk)
train_y, test_y = split_for_train_test(dist_p1[1], msk)
train_1 = [train_x, train_y]
test_1 = [test_x, test_y]
train_x, test_x = split_for_train_test(dist_p2[0], msk)
train_y, test_y = split_for_train_test(dist_p2[1], msk)
train_2 = [train_x, train_y]
test_2 = [test_x, test_y]
u1 = mean(train_1, axis=1)
u2 = mean(train_2, axis=1)
s1 = std(train_1, axis=1)
s2 = std(train_2, axis=1)