def find_best_n_features_mic(n=8, out_path=''):
# 计算MIC
mine = MINE(alpha=0.6, c=15, est="mic_approx")
mic_all = []
for i in range(x.shape[1]):
xi = x[:, i]
mine.compute_score(xi, y)
mic_all.append(mine.mic())
# 找出8个最大的
best_n = []
best_n_mic = []
for i in range(n):
best_position = np.nanargmax(mic_all)
best_n.append(best_position)
best_n_mic.append(copy.deepcopy(mic_all[best_position]))
mic_all[best_position] = np.nan
print('Found', n, 'features with largest MIC, whose positions are:')
print(best_n)
print()
print('The MIC of these features are:')
print(best_n_mic)
print()
best_features = x[:, best_n]
print('Shape of features selected:', best_features.shape)
best_features_with_label = pd.DataFrame(
np.concatenate([best_features, y.reshape(len(y), 1)], axis=1))
out_path = out_path + 'mic_best_' + str(n) + '.csv'
best_features_with_label.to_csv(out_path, header=None, index=None)
def find_best_n_features_relief(n=8, out_path=''):
relief_all = calculate_relief(x, y)
# 找出8个最大的
best_n = []
best_n_ref = []
for i in range(n):
best_position = np.nanargmax(relief_all)
best_n.append(best_position)
best_n_ref.append(copy.deepcopy(relief_all[best_position]))
relief_all[best_position] = np.nan
print('Found', n,
'features with largest Relief Statistics, whose positions are:')
print(best_n)
print()
print('The Relief Statistics of these features are:')
print(best_n_ref)
print()
best_features = x[:, best_n]
print('Shape of features selected:', best_features.shape)
best_features_with_label = pd.DataFrame(
np.concatenate([best_features, y.reshape(len(y), 1)], axis=1))
out_path = out_path + 'relief_best_' + str(n) + '.csv'
best_features_with_label.to_csv(out_path, header=None, index=None)
def find_best_n_features_ridge(n=8, out_path=''):
ridge = LogisticRegression(penalty='l2', class_weight='balanced')
params_tuned = {'C': [1, 2, 3, 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 100, 180, 190,
200, 220, 240, 250, 260, 300, 500, 1000, 2000]}
clf = GridSearchCV(ridge, params_tuned, scoring='accuracy', cv=5)
clf.fit(x, y)
print('During Ridge, the penalty parameter alpha is set as', clf.best_params_['C'])
clf.best_estimator_.fit(x, y)
clf.best_estimator_.coef_ = np.abs(clf.best_estimator_.coef_)
best_n = []
# best_n_coef = []
for i in range(n):
best_position = np.nanargmax(clf.best_estimator_.coef_)
best_n.append(best_position)
# best_n_coef.append(copy.deepcopy(clf.best_estimator_.coef_[:, best_position][0]))
clf.best_estimator_.coef_[:, best_position] = np.nan
print('Selected Features:', best_n)
best_features = x[:, best_n]
print('Shape of features selected:', best_features.shape)
best_features_with_label = pd.DataFrame(np.concatenate([best_features, y.reshape(len(y), 1)], axis=1))
out_path = out_path + 'ridge_best_' + str(n) + '.csv'
best_features_with_label.to_csv(out_path, header=None, index=None)
def find_best_n_features_lasso(n=8, out_path='', max_iter=100):
def find_proper_c(n, low, up, max_iter=100):
c_now = (low + up) / 2
lasso = LogisticRegression(penalty='l1',
class_weight='balanced',
C=c_now)
lasso.fit(x, y)
# 用二分法缩小范围，直到最后留下刚好n个特征
count = 0
fea_num = np.sum(lasso.coef_ > 0.0)
while count < max_iter and (fea_num > n * 1.05 or fea_num < n):
if fea_num > n:
up = (low + up) / 2
elif fea_num < n:
low = (low + up) / 2
c_now = (low + up) / 2
lasso = LogisticRegression(penalty='l1',
class_weight='balanced',
C=c_now)
lasso.fit(x, y)
fea_num = np.sum(lasso.coef_ > 0.0)
count += 1
# print(c_now, ':', fea_num)
return c_now, count
low = 0
up = 100000
c_now, count = find_proper_c(n, low, up, max_iter)
retry = 1
while count == max_iter and retry <= 10:
print('Didn\'t find proper C within', low, 'and', up, ', retrying',
retry)
low = up
up = up * 10
c_now, count = find_proper_c(low, up, max_iter)
retry += 1
if retry == 11:
print('Cound not find proper C, please try a lower dimension.')
return
lasso = LogisticRegression(penalty='l1', class_weight='balanced', C=c_now)
lasso.fit(x, y)
best_features = x[:, (lasso.coef_ > 0.0)[0]]
if best_features.shape[1] > n:
best_n = []
for i in range(n):
best_position = np.nanargmax(lasso.coef_)
best_n.append(best_position)
lasso.coef_[:, best_position] = np.nan
best_features = x[:, best_n]
print('Shape of features selected:', best_features.shape)
best_features_with_label = pd.DataFrame(
np.concatenate([best_features, y.reshape(len(y), 1)], axis=1))
out_path = out_path + 'lasso_best_' + str(n) + '.csv'
best_features_with_label.to_csv(out_path, header=None, index=None)