-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathcreate_meta_dataset.py
128 lines (107 loc) · 5.85 KB
/
create_meta_dataset.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
import json
from pathlib import Path
import pandas as pd
def collect_log_data():
metadata_logs = []
fit_logs = []
predict_logs = []
evaluate_logs = []
for data_set_folder in Path("./").iterdir():
if data_set_folder.is_file():
continue
data_set_name = data_set_folder.name
print(data_set_name)
for algorithm_folder in data_set_folder.iterdir():
if algorithm_folder.is_file():
continue
if "checkpoint" in algorithm_folder.name:
algorithm_name = algorithm_folder.name.split("_")[-1]
for config_folder in algorithm_folder.iterdir():
if config_folder.is_file():
continue
if "config" in config_folder.name:
for fold_folder in config_folder.iterdir():
if fold_folder.is_file():
continue
if "fold" in fold_folder.name:
fold_name = fold_folder.name.split("_")[-1]
if (fold_folder / "fit_log.json").exists():
with open(fold_folder / "fit_log.json", "r") as file:
fit_log = {
"data_set_name": data_set_name,
"algorithm_name": algorithm_name,
"fold_name": fold_name
}
content = json.load(file)
fit_log.update(content)
fit_logs.append(fit_log)
if (fold_folder / "predict_log.json").exists():
with open(fold_folder / "predict_log.json", "r") as file:
predict_log = {
"data_set_name": data_set_name,
"algorithm_name": algorithm_name,
"fold_name": fold_name
}
content = json.load(file)
predict_log.update(content)
predict_logs.append(predict_log)
if (fold_folder / "evaluate_log.json").exists():
with open(fold_folder / "evaluate_log.json", "r") as file:
evaluate_log = {
"data_set_name": data_set_name,
"algorithm_name": algorithm_name,
"fold_name": fold_name
}
content = json.load(file)
evaluate_log.update(content)
evaluate_logs.append(evaluate_log)
if "atomic" in algorithm_folder.name:
if (algorithm_folder / "metadata.json").exists():
with open(algorithm_folder / "metadata.json", "r") as file:
metadata_log = {
"data_set_name": data_set_name
}
content = json.load(file)
metadata_log.update(content)
metadata_logs.append(metadata_log)
metadata = pd.DataFrame(metadata_logs)
fit = pd.DataFrame(fit_logs)
predict = pd.DataFrame(predict_logs)
evaluate = pd.DataFrame(evaluate_logs)
metadata.to_csv("metadata.csv", index=False)
fit.to_csv("fit.csv", index=False)
predict.to_csv("predict.csv", index=False)
evaluate.to_csv("evaluate.csv", index=False)
def merge_log_data():
fit = pd.read_csv("fit.csv")
predict = pd.read_csv("predict.csv")
evaluate = pd.read_csv("evaluate.csv")
merged = pd.merge(predict, evaluate, on=["data_set_name", "algorithm_name", "algorithm_config_index", "fold"],
suffixes=("", "_drop"))
merged.drop([col for col in merged.columns if "_drop" in col], axis=1, inplace=True)
merged = pd.merge(fit, merged, on=["data_set_name", "algorithm_name", "algorithm_config_index", "fold"],
suffixes=("", "_drop"))
merged.drop([col for col in merged.columns if "_drop" in col], axis=1, inplace=True)
merged.drop(columns=["fold_name", "fold", "model_file", "algorithm_configuration"], inplace=True)
merged.to_csv("merged.csv", index=False)
merged = merged.groupby(["data_set_name", "algorithm_name", "algorithm_config_index"]).mean().reset_index()
merged.to_csv("merged_mean.csv", index=False)
def create_metaset():
merged_mean = pd.read_csv("merged_mean.csv")
metadata = pd.read_csv("metadata.csv")
metadata = metadata.set_index("data_set_name")
metadata.drop(columns=["feedback_type"], inplace=True)
metrics = list(merged_mean.columns[-15:])
for metric in metrics:
metric_set = merged_mean[["data_set_name", "algorithm_name", "algorithm_config_index", metric]].copy()
metric_set.loc[:, "algorithm"] = metric_set["algorithm_name"] + "_" + metric_set[
"algorithm_config_index"].astype(str)
metric_set.drop(columns=["algorithm_name", "algorithm_config_index"], inplace=True)
metric_set = metric_set.pivot(index="data_set_name", columns="algorithm", values=metric)
metric_set.dropna(inplace=True)
metaset = pd.merge(metadata, metric_set, left_index=True, right_index=True)
metaset.to_csv(f"metaset_{metric}.csv", index=False)
if __name__ == "__main__":
collect_log_data()
merge_log_data()
create_metaset()