File size: 3,676 Bytes
b265364 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 |
# util functions about data
from scipy.stats import rankdata, iqr, trim_mean
from sklearn.metrics import f1_score, mean_squared_error
import numpy as np
from numpy import percentile
def get_attack_interval(attack):
heads = []
tails = []
for i in range(len(attack)):
if attack[i] == 1:
if attack[i-1] == 0:
heads.append(i)
if i < len(attack)-1 and attack[i+1] == 0:
tails.append(i)
elif i == len(attack)-1:
tails.append(i)
res = []
for i in range(len(heads)):
res.append((heads[i], tails[i]))
# print(heads, tails)
return res
# calculate F1 scores
def eval_scores(scores, true_scores, th_steps, return_thresold=False):
padding_list = [0]*(len(true_scores) - len(scores))
# print(padding_list)
if len(padding_list) > 0:
scores = padding_list + scores
scores_sorted = rankdata(scores, method='ordinal')
th_steps = th_steps
# th_steps = 500
th_vals = np.array(range(th_steps)) * 1.0 / th_steps
fmeas = [None] * th_steps
thresholds = [None] * th_steps
for i in range(th_steps):
cur_pred = scores_sorted > th_vals[i] * len(scores)
fmeas[i] = f1_score(true_scores, cur_pred)
score_index = scores_sorted.tolist().index(int(th_vals[i] * len(scores)+1))
thresholds[i] = scores[score_index]
if return_thresold:
return fmeas, thresholds
return fmeas
def eval_mseloss(predicted, ground_truth):
ground_truth_list = np.array(ground_truth)
predicted_list = np.array(predicted)
# mask = (ground_truth_list == 0) | (predicted_list == 0)
# ground_truth_list = ground_truth_list[~mask]
# predicted_list = predicted_list[~mask]
# neg_mask = predicted_list < 0
# predicted_list[neg_mask] = 0
# err = np.abs(predicted_list / ground_truth_list - 1)
# acc = (1 - np.mean(err))
# return loss
loss = mean_squared_error(predicted_list, ground_truth_list)
return loss
def get_err_median_and_iqr(predicted, groundtruth):
np_arr = np.abs(np.subtract(np.array(predicted), np.array(groundtruth)))
err_median = np.median(np_arr)
err_iqr = iqr(np_arr)
return err_median, err_iqr
def get_err_median_and_quantile(predicted, groundtruth, percentage):
np_arr = np.abs(np.subtract(np.array(predicted), np.array(groundtruth)))
err_median = np.median(np_arr)
# err_iqr = iqr(np_arr)
err_delta = percentile(np_arr, int(percentage*100)) - percentile(np_arr, int((1-percentage)*100))
return err_median, err_delta
def get_err_mean_and_quantile(predicted, groundtruth, percentage):
np_arr = np.abs(np.subtract(np.array(predicted), np.array(groundtruth)))
err_median = trim_mean(np_arr, percentage)
# err_iqr = iqr(np_arr)
err_delta = percentile(np_arr, int(percentage*100)) - percentile(np_arr, int((1-percentage)*100))
return err_median, err_delta
def get_err_mean_and_std(predicted, groundtruth):
np_arr = np.abs(np.subtract(np.array(predicted), np.array(groundtruth)))
err_mean = np.mean(np_arr)
err_std = np.std(np_arr)
return err_mean, err_std
def get_f1_score(scores, gt, contamination):
padding_list = [0]*(len(gt) - len(scores))
# print(padding_list)
threshold = percentile(scores, 100 * (1 - contamination))
if len(padding_list) > 0:
scores = padding_list + scores
pred_labels = (scores > threshold).astype('int').ravel()
return f1_score(gt, pred_labels) |