-
Notifications
You must be signed in to change notification settings - Fork 40
/
metrics_expts.py
189 lines (166 loc) · 8.52 KB
/
metrics_expts.py
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
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
from src.datasets import MultiEntityDataset
from configs import datasets_config, get_thres_config
from src.evaluation.evaluator import *
from src.evaluation.evaluation_utils import *
import matplotlib
matplotlib.rcParams['pdf.fonttype'] = 42
matplotlib.rcParams['ps.fonttype'] = 42
matplotlib.rcParams['mathtext.fontset'] = 'cm'
matplotlib.rcParams['font.family'] = 'STIXGeneral'
def run_random_scoring(root_dir, ds=["damadics-s", "msl", "smap", "smd", "swat", "wadi"]):
os.makedirs(root_dir, exist_ok=True)
init_logging(os.path.join(root_dir, 'logs'))
logger = logging.getLogger(__name__)
seed = 0
np.random.seed(seed)
fscores_dict_raw_t = dict()
fscores_dict_avg_t = dict()
fscores_dict_raw_c = dict()
fscores_dict_avg_c = dict()
fscores_dict_raw_pa = dict()
fscores_dict_avg_pa = dict()
for ds_name in ds:
if ds_name in datasets_config.keys():
ds_kwargs = datasets_config[ds_name]
else:
ds_kwargs = {}
print(ds_name)
ds_class = get_dataset_class(ds_name)
ds_multi = MultiEntityDataset(dataset_class=ds_class, seed=seed, ds_kwargs=ds_kwargs)
thres_config_dict = get_thres_config(ds_name)
fscores_pa = []
fscores_t = []
fscores_c = []
for entity in ds_multi.datasets:
_, _, _, y_test = entity.data()
true_events = get_events(y_test)
# random scores
score_t_test = np.random.rand(len(y_test))
thres_method = "best_f1_test"
test_anom_frac_entity = entity.get_anom_frac_entity()
_, pred_labels = threshold_and_predict(score_t_test, y_test, true_events=true_events, logger=logger,
thres_method=thres_method, point_adjust=True,
score_t_train=None, thres_config_dict=thres_config_dict,
test_anom_frac=test_anom_frac_entity)
print("num predicted by pa {}".format(sum(pred_labels)/len(pred_labels)))
_, _, _, _, _, fscore_pa = evaluate_predicted_labels(pred_labels, y_test, logger=logger,
true_events=true_events,
eval_method="time-wise", point_adjust=True)
_, pred_labels = threshold_and_predict(score_t_test, y_test, true_events=true_events, logger=logger,
thres_method=thres_method, point_adjust=False,
score_t_train=None, thres_config_dict=thres_config_dict,
test_anom_frac=test_anom_frac_entity)
print("num predicted by time-wise f1 {}".format(sum(pred_labels)/len(pred_labels)))
_, _, _, _, _, fscore_t = evaluate_predicted_labels(pred_labels, y_test, logger=logger,
true_events=true_events,
eval_method="time-wise", point_adjust=False)
_, pred_labels = threshold_and_predict(score_t_test, y_test, true_events=true_events,
logger=logger,
thres_method=thres_method, point_adjust=False,
score_t_train=None, thres_config_dict=thres_config_dict,
return_auc=False, composite_best_f1=True,
test_anom_frac=test_anom_frac_entity)
print("num predicted by fc_1 {}".format(sum(pred_labels)/len(pred_labels)))
fscore_c = get_composite_fscore_raw(pred_labels, true_events, y_test)
fscores_pa.append(fscore_pa)
fscores_t.append(fscore_t)
fscores_c.append(fscore_c)
print(fscore_pa)
print(fscore_t)
print(fscore_c)
fscores_dict_raw_pa[ds_name] = fscores_pa
fscores_dict_avg_pa[ds_name] = np.nanmean(fscores_pa)
fscores_dict_raw_t[ds_name] = fscores_t
fscores_dict_avg_t[ds_name] = np.nanmean(fscores_t)
fscores_dict_raw_c[ds_name] = fscores_c
fscores_dict_avg_c[ds_name] = np.nanmean(fscores_c)
print(fscores_dict_avg_pa)
print(fscores_dict_avg_t)
print(fscores_dict_avg_c)
df = pd.DataFrame([fscores_dict_avg_t, fscores_dict_avg_pa, fscores_dict_avg_c])
# df["Point adjust"] = [False, True, False]
df["Metric"] = ["Time-wise F1", "Point-Adjust F1", "Composite F1"]
df.set_index("Metric", inplace=True)
print(df)
print(df.to_latex())
df.to_csv(os.path.join(root_dir, "pt_adj_composite_2ds.csv"))
def run_examples_scores(root_dir):
init_logging(os.path.join(root_dir, 'logs'))
logger = logging.getLogger(__name__)
seed = 0
np.random.seed(seed)
fscores_t = []
fscores_fc = []
fscores_pa = []
len_ts = 150
scores_zero = np.zeros(len_ts)
scores_ones = scores_zero + 1
scores_rand = np.random.choice([0, 1], len_ts)
y_test = np.zeros(len_ts)
y_test[10:30] = 1
y_test[60:70] = 1
scores_1 = np.concatenate([[0, 0, 0, 0, 1] * 20])
scores_2 = np.zeros(len_ts)
scores_2[[11, 12, 50, 65, 68]] = 1
scores_3 = np.zeros(len_ts)
scores_3[15] = 1
scores_3[65] = 1
scores_3[40:50] = 1
scores_3[110:120] = 1
scores_4 = np.zeros(len_ts)
scores_4[12:18] = 1
scores_4[62:70] = 1
scores_5 = np.zeros(len_ts)
scores_5[12:28] = 1
scores = np.stack([scores_rand, scores_ones, scores_3, scores_5, scores_2, scores_4])
true_events = get_events(y_test)
labels = ["Random", "All positives", "$rec_e$=1, high $FP_t$", "$rec_e$=0.5, no $FP_t$", "$rec_e$=1, 1 $FP_t$", "$rec_e$=1, no $FP_t$"]
for pred_num in range(scores.shape[0]):
pred_labels = scores[pred_num, :]
_, _, _, _, _, fscore_pa = evaluate_predicted_labels(pred_labels, y_test, logger=logger,
true_events=true_events,
eval_method="time-wise", point_adjust=True)
_, _, _, _, _, fscore_t = evaluate_predicted_labels(pred_labels, y_test, logger=logger,
true_events=true_events,
eval_method="time-wise", point_adjust=False)
fscore_c = get_composite_fscore_raw(pred_labels, true_events, y_test)
fscores_t.append(fscore_t)
fscores_fc.append(fscore_c)
fscores_pa.append(fscore_pa)
print(fscores_t)
print(fscores_fc)
print(fscores_pa)
df = pd.DataFrame(data=np.stack([fscores_t, fscores_pa, fscores_fc],axis=-1),
columns=["Time-wise F1", "Point-Adjust F1", "Composite F1"])
print(df)
fig, axes = plt.subplots(7, figsize=(4, 3))
normal = np.ma.masked_where(y_test != 1, y_test)
axes[0].plot(range(len_ts), y_test, range(len_ts), normal, linewidth=1)
axes[0].set_xticklabels([])
axes[0].set_yticklabels([])
axes[0].set_ylabel("$y$: Ground truth", rotation=0, labelpad=45, fontsize='large')
axes[0].set_xlim(0, 150)
for pred_num in range(scores.shape[0]):
preds = scores[pred_num, :]
axes[pred_num+1].plot(range(len_ts), preds, linewidth=1)
tp = np.where((y_test == 1) & (preds == 1))[0]
axes[pred_num+1].scatter(tp, np.ones(len(tp)), color='brown', s=5)
axes[pred_num + 1].set_ylim(-0.1, 1.1)
axes[pred_num + 1].set_xlim(0, 150)
axes[pred_num + 1].set_xticklabels([])
axes[pred_num + 1].set_yticklabels([])
axes[pred_num + 1].set_ylabel("$\haty_{}$: {}".format(pred_num+1, labels[pred_num]), rotation=0, labelpad=45,
fontsize='large')
df = df.round(4)
fig.subplots_adjust(wspace=0.2, hspace=0.2)
# plt.tight_layout()
print(df.to_latex())
df.to_csv(os.path.join(root_dir, "metrics_eval_new.csv"))
plt.savefig(os.path.join(root_dir, "example_scores.pdf"), bbox_inches='tight')
if __name__ == "__main__":
print(os.getcwd())
ds = ["msl", "smap", "smd", "damadics-s"]
root_dir = os.path.join(os.getcwd(), "reports", "metrics_expts")
os.makedirs(root_dir, exist_ok=True)
run_random_scoring(root_dir, ds=ds)
run_examples_scores(root_dir)