pred = deepcopy(pred_val)
# if predicted proba >= 0.5, this label is set to 1. if all probas < 0.5, the label with largest proba is set to 1
for i in range(pred.shape[0]):
if len(np.where(pred[i] >= th)[0]) > 0:
pred[i][pred[i] >= th] = 1
pred[i][pred[i] < th] = 0
else:
max_index = np.argmax(pred[i])
pred[i] = 0
pred[i][max_index] = 1
# In[ ]:
acc_val = hamming_score(y_val_issue_single, pred)
p_val, r_val, f1_val = f1(y_val_issue_single, pred)
# acc_val = hamming_score(y_val_reshape, pred)
# p_val, r_val, f1_val = f1(y_val_reshape, pred)
# p_S = precision_S(y_valid_multi_reshape, pred)
# r_S = recall_S(y_valid_multi_reshape, pred)
# f1_S = (2 * p_S * r_S)/(p_S + r_S)
print('Eval====>Th: {}, Acc: {}, P: {}, R: {}, F1: {}'.format(
th, acc_val, p_val, r_val, f1_val))
# print('Eval====>Th: {}, P_S: {}, R_S: {}, F1_S: {}'.format(th, p_S, r_S, f1_S))
# In[ ]:
pred = deepcopy(pred_test)
for th in [0.2, 0.3, 0.4, 0.5, 0.6, 0.7]:
pred = deepcopy(pred_val)
# if predicted proba >= 0.5, this label is set to 1. if all probas < 0.5, the label with largest proba is set to 1
for i in range(pred.shape[0]):
if len(np.where(pred[i] >= th)[0]) > 0:
pred[i][pred[i] >= th] = 1
pred[i][pred[i] < th] = 0
else:
max_index = np.argmax(pred[i])
pred[i] = 0
pred[i][max_index] = 1
# In[ ]:
acc_val = hamming_score(y_val_reshape, pred)
p_val, r_val, f1_val = f1(y_val_reshape, pred)
print('Eval====>Th: {}, Acc: {}, P: {}, R: {}, F1: {}'.format(th, acc_val, p_val, r_val, f1_val))
# In[ ]:
pred = deepcopy(pred_test)
for i in range(pred.shape[0]):
if len(np.where(pred[i] >= th)[0]) > 0:
pred[i][pred[i] >= th] = 1
pred[i][pred[i] < th] = 0
else:
max_index = np.argmax(pred[i])
pred[i] = 0
pred[i][max_index] = 1
for th in [0.2, 0.3, 0.4, 0.5, 0.6, 0.7]:
pred = deepcopy(pred_val)
# if predicted proba >= 0.5, this label is set to 1. if all probas < 0.5, the label with largest proba is set to 1
for i in range(pred.shape[0]):
if len(np.where(pred[i] >= th)[0]) > 0:
pred[i][pred[i] >= th] = 1
pred[i][pred[i] < th] = 0
else:
max_index = np.argmax(pred[i])
pred[i] = 0
pred[i][max_index] = 1
# In[ ]:
acc_val = hamming_score(y_val, pred)
p_val, r_val, f1_val = f1(y_val, pred)
print('Eval====>Th: {}, Acc: {}, P: {}, R: {}, F1: {}'.format(
th, acc_val, p_val, r_val, f1_val))
# In[ ]:
pred = deepcopy(pred_test)
for i in range(pred.shape[0]):
if len(np.where(pred[i] >= th)[0]) > 0:
pred[i][pred[i] >= th] = 1
pred[i][pred[i] < th] = 0
else:
max_index = np.argmax(pred[i])
pred[i] = 0