Spaces:
Running
on
Zero
Running
on
Zero
File size: 19,377 Bytes
1e4c9bc |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 |
import numpy as np
from typing import List, Dict, Tuple, Any, Optional
from dataclasses import dataclass, field
import traceback
from scipy import stats
@dataclass
class CalibrationResult:
"""校準結果結構"""
original_scores: List[float]
calibrated_scores: List[float]
score_mapping: Dict[str, float] # breed -> calibrated_score
calibration_method: str
distribution_stats: Dict[str, float]
quality_metrics: Dict[str, float] = field(default_factory=dict)
@dataclass
class ScoreDistribution:
"""分數分布統計"""
mean: float
std: float
min_score: float
max_score: float
percentile_5: float
percentile_95: float
compression_ratio: float # 分數壓縮比率
effective_range: float # 有效分數範圍
class ScoreCalibrator:
"""
動態分數校準系統
解決分數壓縮問題並保持相對排名
"""
def __init__(self):
"""初始化校準器"""
self.calibration_methods = {
'dynamic_range_mapping': self._dynamic_range_mapping,
'percentile_stretching': self._percentile_stretching,
'gaussian_normalization': self._gaussian_normalization,
'sigmoid_transformation': self._sigmoid_transformation
}
self.quality_thresholds = {
'min_effective_range': 0.3, # 最小有效分數範圍
'max_compression_ratio': 0.2, # 最大允許壓縮比率
'target_distribution_range': (0.45, 0.95) # 目標分布範圍
}
def calibrate_scores(self, breed_scores: List[Tuple[str, float]],
method: str = 'auto') -> CalibrationResult:
"""
校準品種分數
Args:
breed_scores: (breed_name, score) 元組列表
method: 校準方法 ('auto', 'dynamic_range_mapping', 'percentile_stretching', etc.)
Returns:
CalibrationResult: 校準結果
"""
try:
if not breed_scores:
return CalibrationResult(
original_scores=[],
calibrated_scores=[],
score_mapping={},
calibration_method='none',
distribution_stats={}
)
# 提取分數和品種名稱
breeds = [item[0] for item in breed_scores]
original_scores = [item[1] for item in breed_scores]
# 分析原始分數分布
distribution = self._analyze_score_distribution(original_scores)
# 選擇校準方法
if method == 'auto':
method = self._select_calibration_method(distribution)
# 應用校準
calibration_func = self.calibration_methods.get(method, self._dynamic_range_mapping)
calibrated_scores = calibration_func(original_scores, distribution)
# 保持排名一致性
calibrated_scores = self._preserve_ranking(original_scores, calibrated_scores)
# 建立分數映射
score_mapping = dict(zip(breeds, calibrated_scores))
# 計算品質指標
quality_metrics = self._calculate_quality_metrics(
original_scores, calibrated_scores, distribution
)
return CalibrationResult(
original_scores=original_scores,
calibrated_scores=calibrated_scores,
score_mapping=score_mapping,
calibration_method=method,
distribution_stats=self._distribution_to_dict(distribution),
quality_metrics=quality_metrics
)
except Exception as e:
print(f"Error calibrating scores: {str(e)}")
print(traceback.format_exc())
# 回傳原始分數作為降級方案
breeds = [item[0] for item in breed_scores]
original_scores = [item[1] for item in breed_scores]
return CalibrationResult(
original_scores=original_scores,
calibrated_scores=original_scores,
score_mapping=dict(zip(breeds, original_scores)),
calibration_method='fallback',
distribution_stats={}
)
def _analyze_score_distribution(self, scores: List[float]) -> ScoreDistribution:
"""分析分數分布"""
try:
scores_array = np.array(scores)
# 基本統計
mean_score = np.mean(scores_array)
std_score = np.std(scores_array)
min_score = np.min(scores_array)
max_score = np.max(scores_array)
# 百分位數
percentile_5 = np.percentile(scores_array, 5)
percentile_95 = np.percentile(scores_array, 95)
# 壓縮比率和有效範圍
full_range = max_score - min_score
effective_range = percentile_95 - percentile_5
compression_ratio = 1.0 - (effective_range / 1.0) if full_range > 0 else 0.0
return ScoreDistribution(
mean=mean_score,
std=std_score,
min_score=min_score,
max_score=max_score,
percentile_5=percentile_5,
percentile_95=percentile_95,
compression_ratio=compression_ratio,
effective_range=effective_range
)
except Exception as e:
print(f"Error analyzing score distribution: {str(e)}")
# 返回預設分布
return ScoreDistribution(
mean=0.5, std=0.1, min_score=0.0, max_score=1.0,
percentile_5=0.4, percentile_95=0.6,
compression_ratio=0.6, effective_range=0.2
)
def _select_calibration_method(self, distribution: ScoreDistribution) -> str:
"""根據分布特性選擇校準方法"""
# 高度壓縮的分數需要強力展開
if distribution.compression_ratio > 0.8:
return 'percentile_stretching'
# 中等壓縮使用動態範圍映射
elif distribution.compression_ratio > 0.5:
return 'dynamic_range_mapping'
# 分數集中在中間使用 sigmoid 轉換
elif 0.4 <= distribution.mean <= 0.6 and distribution.std < 0.1:
return 'sigmoid_transformation'
# 其他情況使用高斯正規化
else:
return 'gaussian_normalization'
def _dynamic_range_mapping(self, scores: List[float],
distribution: ScoreDistribution) -> List[float]:
"""動態範圍映射校準"""
try:
scores_array = np.array(scores)
# 使用5%和95%百分位數作為邊界
lower_bound = distribution.percentile_5
upper_bound = distribution.percentile_95
# 避免除零
if upper_bound - lower_bound < 0.001:
upper_bound = distribution.max_score
lower_bound = distribution.min_score
if upper_bound - lower_bound < 0.001:
return scores # 所有分數相同,無需校準
# 映射到目標範圍 [0.45, 0.95]
target_min, target_max = self.quality_thresholds['target_distribution_range']
# 線性映射
normalized = (scores_array - lower_bound) / (upper_bound - lower_bound)
normalized = np.clip(normalized, 0, 1) # 限制在 [0,1] 範圍
calibrated = target_min + normalized * (target_max - target_min)
return calibrated.tolist()
except Exception as e:
print(f"Error in dynamic range mapping: {str(e)}")
return scores
def _percentile_stretching(self, scores: List[float],
distribution: ScoreDistribution) -> List[float]:
"""百分位數拉伸校準"""
try:
scores_array = np.array(scores)
# 計算百分位數排名
percentile_ranks = stats.rankdata(scores_array, method='average') / len(scores_array)
# 使用平方根轉換來增強差異
stretched_ranks = np.sqrt(percentile_ranks)
# 映射到目標範圍
target_min, target_max = self.quality_thresholds['target_distribution_range']
calibrated = target_min + stretched_ranks * (target_max - target_min)
return calibrated.tolist()
except Exception as e:
print(f"Error in percentile stretching: {str(e)}")
return self._dynamic_range_mapping(scores, distribution)
def _gaussian_normalization(self, scores: List[float],
distribution: ScoreDistribution) -> List[float]:
"""高斯正規化校準"""
try:
scores_array = np.array(scores)
# Z-score 正規化
if distribution.std > 0:
z_scores = (scores_array - distribution.mean) / distribution.std
# 限制 Z-scores 在合理範圍內
z_scores = np.clip(z_scores, -3, 3)
else:
z_scores = np.zeros_like(scores_array)
# 轉換到目標範圍
target_min, target_max = self.quality_thresholds['target_distribution_range']
target_mean = (target_min + target_max) / 2
target_std = (target_max - target_min) / 6 # 3-sigma 範圍
calibrated = target_mean + z_scores * target_std
calibrated = np.clip(calibrated, target_min, target_max)
return calibrated.tolist()
except Exception as e:
print(f"Error in gaussian normalization: {str(e)}")
return self._dynamic_range_mapping(scores, distribution)
def _sigmoid_transformation(self, scores: List[float],
distribution: ScoreDistribution) -> List[float]:
"""Sigmoid 轉換校準"""
try:
scores_array = np.array(scores)
# 中心化分數
centered = scores_array - distribution.mean
# Sigmoid 轉換 (增強中等分數的差異)
sigmoid_factor = 10.0 # 控制 sigmoid 的陡峭程度
transformed = 1 / (1 + np.exp(-sigmoid_factor * centered))
# 映射到目標範圍
target_min, target_max = self.quality_thresholds['target_distribution_range']
calibrated = target_min + transformed * (target_max - target_min)
return calibrated.tolist()
except Exception as e:
print(f"Error in sigmoid transformation: {str(e)}")
return self._dynamic_range_mapping(scores, distribution)
def _preserve_ranking(self, original_scores: List[float],
calibrated_scores: List[float]) -> List[float]:
"""確保校準後的分數保持原始排名"""
try:
# 獲取原始排名
original_ranks = stats.rankdata([-score for score in original_scores], method='ordinal')
# 獲取校準後的排名
calibrated_with_ranks = list(zip(calibrated_scores, original_ranks))
# 按原始排名排序校準後的分數
calibrated_with_ranks.sort(key=lambda x: x[1])
# 重新分配分數以保持排名但使用校準後的分布
sorted_calibrated = sorted(calibrated_scores, reverse=True)
# 建立新的分數列表
preserved_scores = [0.0] * len(original_scores)
for i, (_, original_rank) in enumerate(calibrated_with_ranks):
# 找到原始位置
original_index = original_ranks.tolist().index(original_rank)
preserved_scores[original_index] = sorted_calibrated[i]
return preserved_scores
except Exception as e:
print(f"Error preserving ranking: {str(e)}")
return calibrated_scores
def _calculate_quality_metrics(self, original_scores: List[float],
calibrated_scores: List[float],
distribution: ScoreDistribution) -> Dict[str, float]:
"""計算校準品質指標"""
try:
original_array = np.array(original_scores)
calibrated_array = np.array(calibrated_scores)
# 範圍改善
original_range = np.max(original_array) - np.min(original_array)
calibrated_range = np.max(calibrated_array) - np.min(calibrated_array)
range_improvement = calibrated_range / max(0.001, original_range)
# 分離度改善 (相鄰分數間的平均差異)
original_sorted = np.sort(original_array)
calibrated_sorted = np.sort(calibrated_array)
original_separation = np.mean(np.diff(original_sorted)) if len(original_sorted) > 1 else 0
calibrated_separation = np.mean(np.diff(calibrated_sorted)) if len(calibrated_sorted) > 1 else 0
separation_improvement = (calibrated_separation / max(0.001, original_separation)
if original_separation > 0 else 1.0)
# 排名保持度 (Spearman 相關係數)
if len(original_scores) > 1:
rank_correlation, _ = stats.spearmanr(original_scores, calibrated_scores)
rank_correlation = abs(rank_correlation) if not np.isnan(rank_correlation) else 1.0
else:
rank_correlation = 1.0
# 分布品質
calibrated_std = np.std(calibrated_array)
distribution_quality = min(1.0, calibrated_std * 2) # 標準差越大品質越好(在合理範圍內)
return {
'range_improvement': range_improvement,
'separation_improvement': separation_improvement,
'rank_preservation': rank_correlation,
'distribution_quality': distribution_quality,
'effective_range_achieved': calibrated_range,
'compression_reduction': max(0, distribution.compression_ratio -
(1.0 - calibrated_range))
}
except Exception as e:
print(f"Error calculating quality metrics: {str(e)}")
return {'error': str(e)}
def _distribution_to_dict(self, distribution: ScoreDistribution) -> Dict[str, float]:
"""將分布統計轉換為字典"""
return {
'mean': distribution.mean,
'std': distribution.std,
'min_score': distribution.min_score,
'max_score': distribution.max_score,
'percentile_5': distribution.percentile_5,
'percentile_95': distribution.percentile_95,
'compression_ratio': distribution.compression_ratio,
'effective_range': distribution.effective_range
}
def apply_tie_breaking(self, breed_scores: List[Tuple[str, float]]) -> List[Tuple[str, float]]:
"""應用確定性的打破平手機制"""
try:
# 按分數分組
score_groups = {}
for breed, score in breed_scores:
rounded_score = round(score, 6) # 避免浮點數精度問題
if rounded_score not in score_groups:
score_groups[rounded_score] = []
score_groups[rounded_score].append((breed, score))
# 處理每個分數組
result = []
for rounded_score in sorted(score_groups.keys(), reverse=True):
group = score_groups[rounded_score]
if len(group) == 1:
result.extend(group)
else:
# 按品種名稱字母順序打破平手
sorted_group = sorted(group, key=lambda x: x[0])
# 為平手的品種分配微小的分數差異
for i, (breed, original_score) in enumerate(sorted_group):
adjusted_score = original_score - (i * 0.0001)
result.append((breed, adjusted_score))
return result
except Exception as e:
print(f"Error in tie breaking: {str(e)}")
return breed_scores
def get_calibration_summary(self, result: CalibrationResult) -> Dict[str, Any]:
"""獲取校準摘要資訊"""
try:
summary = {
'method_used': result.calibration_method,
'breeds_processed': len(result.original_scores),
'score_range_before': {
'min': min(result.original_scores) if result.original_scores else 0,
'max': max(result.original_scores) if result.original_scores else 0,
'range': (max(result.original_scores) - min(result.original_scores))
if result.original_scores else 0
},
'score_range_after': {
'min': min(result.calibrated_scores) if result.calibrated_scores else 0,
'max': max(result.calibrated_scores) if result.calibrated_scores else 0,
'range': (max(result.calibrated_scores) - min(result.calibrated_scores))
if result.calibrated_scores else 0
},
'distribution_stats': result.distribution_stats,
'quality_metrics': result.quality_metrics,
'improvement_summary': {
'range_expanded': result.quality_metrics.get('range_improvement', 1.0) > 1.1,
'separation_improved': result.quality_metrics.get('separation_improvement', 1.0) > 1.1,
'ranking_preserved': result.quality_metrics.get('rank_preservation', 1.0) > 0.95
}
}
return summary
except Exception as e:
print(f"Error generating calibration summary: {str(e)}")
return {'error': str(e)}
def calibrate_breed_scores(breed_scores: List[Tuple[str, float]],
method: str = 'auto') -> CalibrationResult:
"""
便利函數:校準品種分數
Args:
breed_scores: (breed_name, score) 元組列表
method: 校準方法
Returns:
CalibrationResult: 校準結果
"""
calibrator = ScoreCalibrator()
return calibrator.calibrate_scores(breed_scores, method)
def get_calibrated_rankings(breed_scores: List[Tuple[str, float]],
method: str = 'auto') -> List[Tuple[str, float, int]]:
"""
便利函數:獲取校準後的排名
Args:
breed_scores: (breed_name, score) 元組列表
method: 校準方法
Returns:
List[Tuple[str, float, int]]: (breed_name, calibrated_score, rank) 列表
"""
calibrator = ScoreCalibrator()
result = calibrator.calibrate_scores(breed_scores, method)
# 打破平手機制
calibrated_with_breed = [(breed, result.score_mapping[breed]) for breed in result.score_mapping]
calibrated_with_tie_breaking = calibrator.apply_tie_breaking(calibrated_with_breed)
# 添加排名
ranked_results = []
for rank, (breed, score) in enumerate(calibrated_with_tie_breaking, 1):
ranked_results.append((breed, score, rank))
return ranked_results
|