File size: 25,364 Bytes
8986ff6 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 |
"""
ML Model Manager for TIPM
=========================
Centralized management and orchestration of all ML models,
including training, prediction, and model lifecycle management.
"""
import logging
from typing import Dict, List, Optional, Any, Union, Tuple
from datetime import datetime
import json
import pickle
from pathlib import Path
# ML Libraries
try:
import numpy as np
import pandas as pd
NUMPY_AVAILABLE = True
except ImportError:
NUMPY_AVAILABLE = False
np = None
pd = None
# Base classes
from .base import BaseMLModel, ModelType, PredictionResult, ModelStatus
from .classifiers import (
TariffImpactClassifier,
EconomicOutcomeClassifier,
PolicyEffectivenessClassifier,
IndustryVulnerabilityClassifier,
)
from .forecasters import (
GDPImpactForecaster,
TradeFlowForecaster,
EmploymentForecaster,
PriceImpactForecaster,
)
from .ensemble import TIPMEnsemble, ModelVoting, StackingEnsemble, DynamicEnsemble
from .explainability import SHAPExplainer, PolicyInsightGenerator
logger = logging.getLogger(__name__)
class MLModelManager:
"""
Centralized ML model manager for TIPM
Manages all ML models, their training, predictions, and lifecycle.
Provides unified interface for model operations.
"""
def __init__(self, models_dir: str = "models"):
"""
Initialize the ML model manager
Args:
models_dir: Directory to store trained models
"""
self.models_dir = Path(models_dir)
self.models_dir.mkdir(parents=True, exist_ok=True)
# Model registry
self.models: Dict[str, BaseMLModel] = {}
self.model_configs: Dict[str, Dict[str, Any]] = {}
# Model categories
self.classifiers: Dict[str, BaseMLModel] = {}
self.forecasters: Dict[str, BaseMLModel] = {}
self.ensembles: Dict[str, BaseMLModel] = {}
# Explainability components
self.explainers: Dict[str, SHAPExplainer] = {}
self.insight_generators: Dict[str, PolicyInsightGenerator] = {}
# Performance tracking
self.performance_history: Dict[str, List[Dict[str, Any]]] = {}
logger.info(f"Initialized MLModelManager with models directory: {models_dir}")
def register_model(self, model: BaseMLModel, category: str = "general"):
"""
Register a model with the manager
Args:
model: ML model to register
category: Category of the model (classifier, forecaster, ensemble)
"""
try:
model_id = model.model_id
if model_id in self.models:
logger.warning(f"Model {model_id} already registered, overwriting")
# Register model
self.models[model_id] = model
# Categorize model
if category == "classifier" or model.model_type in [
ModelType.CLASSIFICATION,
ModelType.MULTI_CLASS,
ModelType.BINARY,
]:
self.classifiers[model_id] = model
elif category == "forecaster" or model.model_type == ModelType.TIME_SERIES:
self.forecasters[model_id] = model
elif category == "ensemble" or model.model_type == ModelType.ENSEMBLE:
self.ensembles[model_id] = model
# Initialize explainability components
if hasattr(model, "predict"):
self.explainers[model_id] = SHAPExplainer(model)
self.insight_generators[model_id] = PolicyInsightGenerator(model)
# Initialize performance tracking
self.performance_history[model_id] = []
logger.info(f"Registered model {model_id} in category {category}")
except Exception as e:
logger.error(f"Failed to register model {model.model_id}: {e}")
raise
def create_default_models(self):
"""Create and register default TIPM models"""
try:
logger.info("Creating default TIPM models...")
# Create classifiers
classifiers = [
TariffImpactClassifier(),
EconomicOutcomeClassifier(),
PolicyEffectivenessClassifier(),
IndustryVulnerabilityClassifier(),
]
for classifier in classifiers:
self.register_model(classifier, "classifier")
# Create forecasters
forecasters = [
GDPImpactForecaster(),
TradeFlowForecaster(),
EmploymentForecaster(),
PriceImpactForecaster(),
]
for forecaster in forecasters:
self.register_model(forecaster, "forecaster")
# Create ensemble
ensemble = TIPMEnsemble()
self.register_model(ensemble, "ensemble")
logger.info(
f"Created {len(classifiers)} classifiers, {len(forecasters)} forecasters, and 1 ensemble"
)
except Exception as e:
logger.error(f"Failed to create default models: {e}")
raise
def train_model(
self,
model_id: str,
X: Union[pd.DataFrame, np.ndarray],
y: Union[pd.Series, np.ndarray],
**kwargs,
) -> bool:
"""
Train a specific model
Args:
model_id: ID of the model to train
X: Training features
y: Training targets
**kwargs: Additional training parameters
Returns:
True if training successful, False otherwise
"""
try:
if model_id not in self.models:
logger.error(f"Model {model_id} not found")
return False
model = self.models[model_id]
logger.info(f"Starting training for model {model_id}")
# Train the model
trained_model = model.fit(X, y, **kwargs)
if trained_model._is_trained:
logger.info(f"Successfully trained model {model_id}")
# Save model
self.save_model(model_id)
# Update performance history
self._update_performance_history(
model_id,
"training",
{
"training_score": model.metadata.training_score,
"feature_count": model.metadata.feature_count,
"sample_count": model.metadata.sample_count,
},
)
return True
else:
logger.error(f"Model {model_id} training failed")
return False
except Exception as e:
logger.error(f"Failed to train model {model_id}: {e}")
return False
def train_all_models(
self,
X: Union[pd.DataFrame, np.ndarray],
y: Union[pd.Series, np.ndarray],
**kwargs,
) -> Dict[str, bool]:
"""
Train all registered models
Args:
X: Training features
y: Training targets
**kwargs: Additional training parameters
Returns:
Dictionary mapping model IDs to training success status
"""
results = {}
for model_id in self.models:
logger.info(f"Training model {model_id}")
success = self.train_model(model_id, X, y, **kwargs)
results[model_id] = success
return results
def predict(
self, model_id: str, X: Union[pd.DataFrame, np.ndarray]
) -> Optional[PredictionResult]:
"""
Make predictions using a specific model
Args:
model_id: ID of the model to use
X: Input features
Returns:
Prediction result or None if failed
"""
try:
if model_id not in self.models:
logger.error(f"Model {model_id} not found")
return None
model = self.models[model_id]
if not model._is_trained:
logger.error(f"Model {model_id} is not trained")
return None
# Make prediction
prediction = model.predict(X)
# Update performance history
self._update_performance_history(
model_id,
"prediction",
{
"processing_time": prediction.processing_time,
"prediction_count": (
len(prediction.predictions)
if hasattr(prediction.predictions, "__len__")
else 1
),
},
)
return prediction
except Exception as e:
logger.error(f"Failed to make prediction with model {model_id}: {e}")
return None
def predict_with_ensemble(
self, X: Union[pd.DataFrame, np.ndarray]
) -> Dict[str, PredictionResult]:
"""
Make predictions using all models and ensemble
Args:
X: Input features
Returns:
Dictionary mapping model IDs to prediction results
"""
results = {}
# Get predictions from individual models
for model_id in self.models:
if model_id not in self.ensembles: # Skip ensemble models
prediction = self.predict(model_id, X)
if prediction:
results[model_id] = prediction
# Get ensemble prediction if available
for ensemble_id in self.ensembles:
prediction = self.predict(ensemble_id, X)
if prediction:
results[ensemble_id] = prediction
return results
def explain_prediction(
self, model_id: str, X: Union[pd.DataFrame, np.ndarray]
) -> Optional[Dict[str, Any]]:
"""
Explain a prediction using SHAP
Args:
model_id: ID of the model to explain
X: Input features for explanation
Returns:
SHAP explanation or None if failed
"""
try:
if model_id not in self.explainers:
logger.error(f"No explainer available for model {model_id}")
return None
explainer = self.explainers[model_id]
# Fit explainer if not already fitted
if explainer.explainer is None:
explainer.fit_explainer(X)
# Generate explanation
explanation = explainer.explain_prediction(X)
return explanation
except Exception as e:
logger.error(f"Failed to explain prediction for model {model_id}: {e}")
return None
def generate_policy_insights(
self, model_id: str, explanation: Dict[str, Any], context: Dict[str, Any] = None
) -> Optional[Dict[str, Any]]:
"""
Generate policy insights from model explanation
Args:
model_id: ID of the model
explanation: SHAP explanation
context: Additional context for insight generation
Returns:
Policy insights or None if failed
"""
try:
if model_id not in self.insight_generators:
logger.error(f"No insight generator available for model {model_id}")
return None
insight_generator = self.insight_generators[model_id]
insights = insight_generator.generate_policy_insights(explanation, context)
return insights
except Exception as e:
logger.error(
f"Failed to generate policy insights for model {model_id}: {e}"
)
return None
def get_model_status(self, model_id: str) -> Optional[Dict[str, Any]]:
"""
Get status information for a specific model
Args:
model_id: ID of the model
Returns:
Model status information or None if not found
"""
if model_id not in self.models:
return None
model = self.models[model_id]
status = {
"model_id": model_id,
"name": model.name,
"description": model.description,
"model_type": model.model_type.value,
"is_trained": model._is_trained,
"status": (
model.metadata.status.value if model.metadata.status else "unknown"
),
"training_score": model.metadata.training_score,
"validation_score": model.metadata.validation_score,
"last_trained": (
model.metadata.last_trained.isoformat()
if model.metadata.last_trained
else None
),
"feature_count": model.metadata.feature_count,
"sample_count": model.metadata.sample_count,
}
return status
def get_all_model_status(self) -> Dict[str, Dict[str, Any]]:
"""
Get status information for all models
Returns:
Dictionary mapping model IDs to status information
"""
status = {}
for model_id in self.models:
model_status = self.get_model_status(model_id)
if model_status:
status[model_id] = model_status
return status
def save_model(self, model_id: str) -> bool:
"""
Save a trained model to disk
Args:
model_id: ID of the model to save
Returns:
True if save successful, False otherwise
"""
try:
if model_id not in self.models:
logger.error(f"Model {model_id} not found")
return False
model = self.models[model_id]
if not model._is_trained:
logger.warning(f"Model {model_id} is not trained, skipping save")
return False
# Create model directory
model_dir = self.models_dir / model_id
model_dir.mkdir(exist_ok=True)
# Save model file
model_path = model_dir / "model.pkl"
with open(model_path, "wb") as f:
pickle.dump(model, f)
# Save metadata
metadata_path = model_dir / "metadata.json"
metadata = {
"model_id": model_id,
"name": model.name,
"description": model.description,
"model_type": model.model_type.value,
"training_score": model.metadata.training_score,
"validation_score": model.metadata.validation_score,
"last_trained": (
model.metadata.last_trained.isoformat()
if model.metadata.last_trained
else None
),
"feature_count": model.metadata.feature_count,
"sample_count": model.metadata.sample_count,
"saved_at": datetime.now().isoformat(),
}
with open(metadata_path, "w") as f:
json.dump(metadata, f, indent=2)
logger.info(f"Saved model {model_id} to {model_path}")
return True
except Exception as e:
logger.error(f"Failed to save model {model_id}: {e}")
return False
def load_model(self, model_id: str) -> bool:
"""
Load a saved model from disk
Args:
model_id: ID of the model to load
Returns:
True if load successful, False otherwise
"""
try:
model_dir = self.models_dir / model_id
model_path = model_dir / "model.pkl"
metadata_path = model_dir / "metadata.json"
if not model_path.exists():
logger.error(f"Model file not found: {model_path}")
return False
# Load model
with open(model_path, "rb") as f:
model = pickle.load(f)
# Verify metadata
if metadata_path.exists():
with open(metadata_path, "r") as f:
metadata = json.load(f)
# Update model metadata
if hasattr(model, "metadata"):
model.metadata.last_trained = datetime.fromisoformat(
metadata.get("last_trained", "")
)
model.metadata.training_score = metadata.get("training_score", 0.0)
model.metadata.validation_score = metadata.get(
"validation_score", 0.0
)
model.metadata.feature_count = metadata.get("feature_count", 0)
model.metadata.sample_count = metadata.get("sample_count", 0)
# Register loaded model
self.register_model(model)
logger.info(f"Loaded model {model_id} from {model_path}")
return True
except Exception as e:
logger.error(f"Failed to load model {model_id}: {e}")
return False
def save_all_models(self) -> Dict[str, bool]:
"""
Save all trained models to disk
Returns:
Dictionary mapping model IDs to save success status
"""
results = {}
for model_id in self.models:
success = self.save_model(model_id)
results[model_id] = success
return results
def load_all_models(self) -> Dict[str, bool]:
"""
Load all saved models from disk
Returns:
Dictionary mapping model IDs to load success status
"""
results = {}
# Find all model directories
for model_dir in self.models_dir.iterdir():
if model_dir.is_dir() and (model_dir / "model.pkl").exists():
model_id = model_dir.name
success = self.load_model(model_id)
results[model_id] = success
return results
def delete_model(self, model_id: str) -> bool:
"""
Delete a model and its files
Args:
model_id: ID of the model to delete
Returns:
True if deletion successful, False otherwise
"""
try:
# Remove from registries
if model_id in self.models:
del self.models[model_id]
if model_id in self.classifiers:
del self.classifiers[model_id]
if model_id in self.forecasters:
del self.forecasters[model_id]
if model_id in self.ensembles:
del self.ensembles[model_id]
if model_id in self.explainers:
del self.explainers[model_id]
if model_id in self.insight_generators:
del self.insight_generators[model_id]
if model_id in self.performance_history:
del self.performance_history[model_id]
# Remove files
model_dir = self.models_dir / model_id
if model_dir.exists():
import shutil
shutil.rmtree(model_dir)
logger.info(f"Deleted model {model_id}")
return True
except Exception as e:
logger.error(f"Failed to delete model {model_id}: {e}")
return False
def get_model_performance(self, model_id: str) -> List[Dict[str, Any]]:
"""
Get performance history for a specific model
Args:
model_id: ID of the model
Returns:
List of performance records
"""
return self.performance_history.get(model_id, [])
def get_overall_performance(self) -> Dict[str, Any]:
"""
Get overall performance summary for all models
Returns:
Overall performance summary
"""
summary = {
"total_models": len(self.models),
"trained_models": sum(1 for m in self.models.values() if m._is_trained),
"model_categories": {
"classifiers": len(self.classifiers),
"forecasters": len(self.forecasters),
"ensembles": len(self.ensembles),
},
"performance_summary": {},
}
# Calculate performance metrics for each model
for model_id in self.models:
model = self.models[model_id]
if model._is_trained:
performance = {
"training_score": model.metadata.training_score or 0.0,
"validation_score": model.metadata.validation_score or 0.0,
"feature_count": model.metadata.feature_count or 0,
"sample_count": model.metadata.sample_count or 0,
}
summary["performance_summary"][model_id] = performance
return summary
def _update_performance_history(
self, model_id: str, operation: str, metrics: Dict[str, Any]
):
"""Update performance history for a model"""
if model_id not in self.performance_history:
self.performance_history[model_id] = []
record = {
"timestamp": datetime.now().isoformat(),
"operation": operation,
"metrics": metrics,
}
self.performance_history[model_id].append(record)
# Keep only last 100 records
if len(self.performance_history[model_id]) > 100:
self.performance_history[model_id] = self.performance_history[model_id][
-100:
]
def export_model_report(self, model_id: str, output_path: str) -> bool:
"""
Export a comprehensive report for a model
Args:
model_id: ID of the model
output_path: Path to save the report
Returns:
True if export successful, False otherwise
"""
try:
if model_id not in self.models:
logger.error(f"Model {model_id} not found")
return False
model = self.models[model_id]
# Gather report data
report = {
"model_info": self.get_model_status(model_id),
"performance_history": self.get_model_performance(model_id),
"training_history": (
model.training_history if hasattr(model, "training_history") else []
),
"hyperparameters": (
model.hyperparameters if hasattr(model, "hyperparameters") else {}
),
"export_timestamp": datetime.now().isoformat(),
}
# Save report
with open(output_path, "w") as f:
json.dump(report, f, indent=2)
logger.info(f"Exported model report for {model_id} to {output_path}")
return True
except Exception as e:
logger.error(f"Failed to export model report for {model_id}: {e}")
return False
def create_ensemble_from_models(
self, ensemble_id: str, model_ids: List[str], ensemble_type: str = "voting"
) -> bool:
"""
Create an ensemble from existing models
Args:
ensemble_id: ID for the new ensemble
model_ids: List of model IDs to include in ensemble
ensemble_type: Type of ensemble (voting, stacking, dynamic)
Returns:
True if ensemble creation successful, False otherwise
"""
try:
# Verify all models exist and are trained
for model_id in model_ids:
if model_id not in self.models:
logger.error(f"Model {model_id} not found")
return False
if not self.models[model_id]._is_trained:
logger.error(f"Model {model_id} is not trained")
return False
# Create appropriate ensemble
if ensemble_type == "voting":
ensemble = ModelVoting(ensemble_id)
for model_id in model_ids:
ensemble.add_model(self.models[model_id])
elif ensemble_type == "stacking":
ensemble = StackingEnsemble(ensemble_id)
for model_id in model_ids:
ensemble.add_base_model(self.models[model_id])
# Use first model as meta-learner (simplified approach)
if model_ids:
ensemble.set_meta_model(self.models[model_ids[0]])
elif ensemble_type == "dynamic":
ensemble = DynamicEnsemble(ensemble_id)
for model_id in model_ids:
ensemble.add_model(self.models[model_id])
else:
logger.error(f"Unsupported ensemble type: {ensemble_type}")
return False
# Register ensemble
self.register_model(ensemble, "ensemble")
logger.info(
f"Created {ensemble_type} ensemble {ensemble_id} with {len(model_ids)} models"
)
return True
except Exception as e:
logger.error(f"Failed to create ensemble {ensemble_id}: {e}")
return False
|