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
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
|
# -*- coding: utf-8 -*-
"""Compare module."""
from __future__ import division
from typing import Dict, Optional, Union
from .errors import pycmCompareError
from .output import *
from .utils import *
from .params import *
from .cm import ConfusionMatrix
import os
import numpy
from warnings import warn
class Compare():
"""
Compare class.
>>> cm1 = ConfusionMatrix(matrix={0:{0:2,1:50,2:6}, 1:{0:5,1:50,2:3}, 2:{0:1,1:7,2:50}})
>>> cm2 = ConfusionMatrix(matrix={0:{0:50,1:2,2:6}, 1:{0:50,1:5,2:3}, 2:{0:1,1:55,2:2}})
>>> cp = Compare({"cm1": cm1, "cm2": cm2})
>>> print(cp)
Best : cm1
<BLANKLINE>
Rank Name Class-Score Overall-Score
1 cm1 0.50278 0.58095
2 cm2 0.33611 0.52857
<BLANKLINE>
>>> cp.best
pycm.ConfusionMatrix(classes: [0, 1, 2])
>>> cp.sorted
['cm1', 'cm2']
>>> cp.best_name
'cm1'
"""
def __init__(
self,
cm_dict: Dict[str, "ConfusionMatrix"],
by_class: bool=False,
class_weight: Optional[dict]=None,
class_benchmark_weight: Optional[dict]=None,
overall_benchmark_weight: Optional[dict]=None,
digit: int=5) -> None:
"""
Init method.
:param cm_dict: dictionary of confusion matrices
:param by_class: compare by class flag
:param class_weight: class weights
:param class_benchmark_weight: class benchmark weights
:param overall_benchmark_weight: overall benchmark weights
:param digit: scale (number of fraction digits)(default value: 5)
"""
self.scores = None
self.sorted = None
self.classes = None
__compare_assign_handler__(
self,
cm_dict,
class_weight,
class_benchmark_weight,
overall_benchmark_weight,
digit)
__compare_class_handler__(self, cm_dict)
__compare_overall_handler__(self, cm_dict)
__compare_rounder__(self, cm_dict)
scores_list = list(self.scores.values())
(max_overall_name, max_class_name) = __compare_sort_handler__(self)
if scores_list.count(self.scores[max_class_name]) == 1:
if by_class:
self.best = cm_dict[max_class_name]
self.best_name = max_class_name
else:
if max_overall_name == max_class_name:
self.best = cm_dict[max_class_name]
self.best_name = max_overall_name
else:
warn(COMPARE_RESULT_WARNING, RuntimeWarning)
else:
warn(COMPARE_RESULT_WARNING, RuntimeWarning)
def print_report(self) -> None:
"""Print Compare report."""
report = compare_report_print(
self.sorted, self.scores, self.best_name)
print(report)
def save_report(
self,
name: str,
address: bool=True) -> Dict[str, Union[bool, str]]:
"""
Save Compare report in .comp (flat file format).
:param name: filename
:param address: flag for address return
"""
try:
message = None
file = open(name + ".comp", "w")
report = compare_report_print(
self.sorted, self.scores, self.best_name)
file.write(report)
file.close()
if address:
message = os.path.join(
os.getcwd(), name + ".comp") # pragma: no cover
return {"Status": True, "Message": message}
except Exception as e:
return {"Status": False, "Message": str(e)}
def __repr__(self) -> str:
"""Compare object representation method."""
return "pycm.Compare(classes: " + str(self.classes) + ")"
def __str__(self) -> str:
"""Compare object string representation method."""
report = compare_report_print(
self.sorted, self.scores, self.best_name)
return report
def __compare_class_handler__(compare: "Compare", cm_dict: Dict[str, "ConfusionMatrix"]) -> None:
"""
Handle class score of Compare class.
:param compare: Compare
:param cm_dict: dictionary of confusion matrices
"""
class_weight_sum = sum(compare.class_weight.values())
class_benchmark_weight_sum = sum(compare.class_benchmark_weight.values())
for c in compare.classes:
for item in CLASS_BENCHMARK_SCORE_DICT:
max_item_score = len(CLASS_BENCHMARK_SCORE_DICT[item]) - 1
all_class_score = [CLASS_BENCHMARK_SCORE_DICT[item][
cm.class_stat[item][c]] for cm in cm_dict.values()]
if all([isinstance(x, int) for x in all_class_score]):
for cm_name in cm_dict:
score = (compare.class_weight[c] / class_weight_sum) * (
CLASS_BENCHMARK_SCORE_DICT[item][cm_dict[cm_name].class_stat[item][c]] / max_item_score)
score = score * \
(compare.class_benchmark_weight[item] / class_benchmark_weight_sum)
compare.scores[cm_name]["class"] += score
def __compare_overall_handler__(compare: "Compare", cm_dict: Dict[str, "ConfusionMatrix"]) -> None:
"""
Handle overall score of Compare class.
:param compare: Compare
:param cm_dict: dictionary of confusion matrices
"""
overall_benchmark_weight_sum = sum(
compare.overall_benchmark_weight.values())
for item in OVERALL_BENCHMARK_SCORE_DICT:
max_item_score = len(OVERALL_BENCHMARK_SCORE_DICT[item]) - 1
all_overall_score = [OVERALL_BENCHMARK_SCORE_DICT[item][
cm.overall_stat[OVERALL_BENCHMARK_MAP[item]]] for cm in cm_dict.values()]
if all([isinstance(x, int) for x in all_overall_score]):
for cm_name in cm_dict:
score = OVERALL_BENCHMARK_SCORE_DICT[item][cm_dict[cm_name]
.overall_stat[OVERALL_BENCHMARK_MAP[item]]] / max_item_score
score = score * \
(compare.overall_benchmark_weight[item] / overall_benchmark_weight_sum)
compare.scores[cm_name]["overall"] += score
def __compare_rounder__(compare: "Compare", cm_dict: Dict[str, "ConfusionMatrix"]) -> None:
"""
Round Compare.scores .
:param compare: Compare object
:param cm_dict: dictionary of confusion matrices
"""
for cm_name in cm_dict:
compare.scores[cm_name]["overall"] = numpy.around(
compare.scores[cm_name]["overall"], compare.digit)
compare.scores[cm_name]["class"] = numpy.around(
compare.scores[cm_name]["class"], compare.digit)
def __compare_sort_handler__(compare: "Compare") -> Tuple[str, str]:
"""
Handle sorting of scores.
:param compare: Compare
"""
sorted_by_class = sorted(
compare.scores,
key=lambda x: (
compare.scores[x]['class'],
compare.scores[x]['overall']))
sorted_by_overall = sorted(
compare.scores,
key=lambda x: (
compare.scores[x]['overall'],
compare.scores[x]['class']))
sorted_by_class.reverse()
sorted_by_overall.reverse()
compare.sorted = sorted_by_class
max_overall_name = sorted_by_overall[0]
max_class_name = sorted_by_class[0]
return (max_overall_name, max_class_name)
def __compare_weight_handler__(compare: "Compare", weight: Dict[str, float], weight_type: str) -> None:
"""
Handle different weights validation.
:param compare: Compare
:param weight: input weight
:param weight_type: input weight type
"""
valid_dict = {
"class_weight": compare.classes,
"class_benchmark_weight": CLASS_BENCHMARK_SCORE_DICT.keys(),
"overall_benchmark_weight": OVERALL_BENCHMARK_SCORE_DICT.keys()}
error_dict = {
"class_weight": COMPARE_CLASS_WEIGHT_ERROR,
"class_benchmark_weight": COMPARE_CLASS_BENCHMARK_WEIGHT_ERROR,
"overall_benchmark_weight": COMPARE_OVERALL_BENCHMARK_WEIGHT_ERROR}
warning_dict = {
"class_weight": COMPARE_CLASS_WEIGHT_WARNING,
"class_benchmark_weight": COMPARE_CLASS_BENCHMARK_WEIGHT_WARNING,
"overall_benchmark_weight": COMPARE_OVERALL_BENCHMARK_WEIGHT_WARNING}
if weight is None:
return None
if not isinstance(weight, dict):
raise pycmCompareError(error_dict[weight_type])
if set(weight) == set(valid_dict[weight_type]):
if all([isfloat(x) for x in weight.values()]
) and sum(weight.values()) != 0:
setattr(compare, weight_type, weight)
else:
warn(warning_dict[weight_type], RuntimeWarning)
else:
raise pycmCompareError(error_dict[weight_type])
def __compare_assign_handler__(
compare: "Compare",
cm_dict: Dict[str, "ConfusionMatrix"],
class_weight: Dict[str, float],
class_benchmark_weight: Dict[str, float],
overall_benchmark_weight: Dict[str, float],
digit: int) -> None:
"""
Assign basic parameters to Compare.
:param compare: Compare
:param cm_dict: dictionary of confusion matrices
:param class_weight: class weights
:param class_benchmark_weight: class benchmark weights
:param overall_benchmark_weight: overall benchmark weights
:param digit: scale (number of fraction digits)(default value: 5)
"""
if not isinstance(cm_dict, dict):
raise pycmCompareError(COMPARE_FORMAT_ERROR)
if not all(isinstance(item, ConfusionMatrix)
for item in cm_dict.values()):
raise pycmCompareError(COMPARE_TYPE_ERROR)
if any(item.metrics_off for item in cm_dict.values()):
raise pycmCompareError(COMPARE_METRICS_OFF_ERROR)
if not list_check_equal([getattr(item, "POP")
for item in cm_dict.values()]):
raise pycmCompareError(COMPARE_DOMAIN_ERROR)
if len(cm_dict) < 2:
raise pycmCompareError(COMPARE_NUMBER_ERROR)
compare.classes = list(cm_dict.values())[0].classes
compare.class_weight = {k: 1 for k in compare.classes}
compare.class_benchmark_weight = {k: 1 for k in CLASS_BENCHMARK_LIST}
compare.overall_benchmark_weight = {
k: 0 if k in KAPPA_BENCHMARK_LIST[1:] else 1 for k in OVERALL_BENCHMARK_LIST}
compare.digit = digit
compare.best = None
compare.best_name = None
compare.sorted = None
compare.scores = {k: {"overall": 0, "class": 0}.copy()
for k in cm_dict}
__compare_weight_handler__(compare, class_weight, "class_weight")
__compare_weight_handler__(
compare,
class_benchmark_weight,
"class_benchmark_weight")
__compare_weight_handler__(
compare,
overall_benchmark_weight,
"overall_benchmark_weight")
|
