-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathevaluation.py
126 lines (101 loc) · 7.16 KB
/
evaluation.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
import numpy as np
import pandas as pd
import logging
from tqdm import tqdm
import datetime
from simplifications import get_OS_simplification, get_RDP_simplification, get_bottom_up_simplification, \
get_VC_simplification
from Utils.metrics import calculate_mean_loyalty, calculate_kappa_loyalty, calculate_complexity
from Utils.load_models import model_batch_classify # type: ignore
logging.basicConfig(level=
logging.debug)
def score_different_alphas(dataset_name, datset_type, model_path):
"""
Evaluate the impact of different alpha values on loyalty, kappa, and complexity.
"""
diff_alpha_values = np.arange(0,1,0.01)
df = pd.DataFrame(columns=["Type","Alpha", "Mean Loyalty", "Kappa Loyalty", "Complexity"])
time_os = []
time_rdp = []
time_vc = []
time_bu = []
for alpha in tqdm(diff_alpha_values):
# Step 1 gen all simplified ts
logging.debug(f"Alpha: {alpha}")
logging.debug("Running OS")
init_time = datetime.datetime.now()
all_time_series_OS, all_simplificationsOS = get_OS_simplification(dataset_name=dataset_name,datset_type=datset_type, alpha=alpha)
time_os.append((datetime.datetime.now()-init_time).total_seconds())
#Step 2 get model predictions
batch_simplified_ts = [ts.line_version for ts in all_simplificationsOS]
predicted_classes_simplifications_OS = get_model_predictions(model_path, batch_simplified_ts)
predicted_classes_original = get_model_predictions(model_path, all_time_series_OS)
# Step 3 calculate loyalty and complexity
mean_loyaltyOS = calculate_mean_loyalty(pred_class_original=predicted_classes_original, pred_class_simplified=predicted_classes_simplifications_OS)
kappa_loyaltyOS = calculate_kappa_loyalty(pred_class_original=predicted_classes_original, pred_class_simplified=predicted_classes_simplifications_OS)
complexityOS = calculate_complexity(batch_simplified_ts=all_simplificationsOS)
row = ["OS", alpha, mean_loyaltyOS, kappa_loyaltyOS, complexityOS]
df.loc[len(df)] = row
# Step 1 gen all simplified ts
logging.debug("Running RDP")
init_time = datetime.datetime.now()
all_time_series_RDP, all_simplifications_RDP = get_RDP_simplification(dataset_name=dataset_name, datset_type=datset_type, epsilon=alpha)
time_rdp.append((datetime.datetime.now()-init_time).total_seconds())
#Step 2 get model predictions
batch_simplified_ts = [ts.line_version for ts in all_simplifications_RDP]
predicted_classes_simplifications_RDP = get_model_predictions(model_path, batch_simplified_ts)
predicted_classes_original = get_model_predictions(model_path, all_time_series_RDP) #I will say this and all_time_series_OS are the same, but just in case
# Step 3 calculate loyalty and complexity
mean_loyalty_RDP = calculate_mean_loyalty(pred_class_original=predicted_classes_original, pred_class_simplified=predicted_classes_simplifications_RDP)
kappa_loyalty_RDP = calculate_kappa_loyalty(pred_class_original=predicted_classes_original, pred_class_simplified=predicted_classes_simplifications_RDP)
complexity_RDP = calculate_complexity(batch_simplified_ts=all_simplifications_RDP)
row = ["RDP", alpha, mean_loyalty_RDP, kappa_loyalty_RDP, complexity_RDP]
df.loc[len(df)] = row
# Step 1 gen all simplified ts
logging.debug("Running BU")
init_time = datetime.datetime.now()
all_time_series_BU, all_simplifications_BU = get_bottom_up_simplification(dataset_name=dataset_name,
datset_type=datset_type, max_error=alpha)
time_bu.append((datetime.datetime.now()-init_time).total_seconds())
# Step 2 get model predictions
batch_simplified_ts = [ts.line_version for ts in all_simplifications_BU]
predicted_classes_simplifications_BU = get_model_predictions(model_path, batch_simplified_ts)
predicted_classes_original = get_model_predictions(model_path,
all_time_series_BU) # I will say this and all_time_series_OS are the same, but just in case
# Step 3 calculate loyalty and complexity
mean_loyalty_BU = calculate_mean_loyalty(pred_class_original=predicted_classes_original,
pred_class_simplified=predicted_classes_simplifications_BU)
kappa_loyalty_BU = calculate_kappa_loyalty(pred_class_original=predicted_classes_original,
pred_class_simplified=predicted_classes_simplifications_BU)
complexity_BU = calculate_complexity(batch_simplified_ts=all_simplifications_BU)
row = ["BU", alpha, mean_loyalty_BU, kappa_loyalty_BU, complexity_BU]
df.loc[len(df)] = row
# Step 1 gen all simplified ts
logging.debug("Running VC")
init_time = datetime.datetime.now()
all_time_series_VC, all_simplifications_VC = get_VC_simplification(dataset_name=dataset_name,
datset_type=datset_type,
alpha=alpha)
time_vc.append((datetime.datetime.now()-init_time).total_seconds())
# Step 2 get model predictions
batch_simplified_ts = [ts.line_version for ts in all_simplifications_VC]
predicted_classes_simplifications_VC = get_model_predictions(model_path, batch_simplified_ts)
predicted_classes_original = get_model_predictions(model_path,
all_time_series_VC) # I will say this and all_time_series_OS are the same, but just in case
# Step 3 calculate loyalty and complexity
mean_loyalty_VC = calculate_mean_loyalty(pred_class_original=predicted_classes_original,
pred_class_simplified=predicted_classes_simplifications_VC)
kappa_loyalty_VC = calculate_kappa_loyalty(pred_class_original=predicted_classes_original,
pred_class_simplified=predicted_classes_simplifications_VC)
complexity_VC = calculate_complexity(batch_simplified_ts=all_simplifications_VC)
row = ["VC", alpha, mean_loyalty_VC, kappa_loyalty_VC, complexity_VC]
df.loc[len(df)] = row
time_log = pd.read_csv(f"results/time_log.csv", header=0)
if time_log.query(f"dataset == '{dataset_name + datset_type}' & model == '{model_path.split('/')[-1]}'") is not None:
time_log = time_log.drop(time_log.query(f"dataset == '{dataset_name + datset_type}' & model == '{model_path.split('/')[-1]}'").index)
time_log.loc[len(time_log)] = [dataset_name + datset_type, model_path.split("/")[-1], np.mean(time_os), np.mean(time_rdp), np.mean(time_vc), np.mean(time_bu)]
time_log.to_csv(f"results/time_log.csv", index=False)
return df
def get_model_predictions(model_path, batch_of_TS):
predicted_classes = model_batch_classify(model_path, batch_of_timeseries=batch_of_TS)
return predicted_classes