File size: 44,417 Bytes
c49b21b |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 |
import pandas as pd
import numpy as np
from sklearn.impute import KNNImputer
from sklearn.preprocessing import StandardScaler
import warnings
warnings.filterwarnings('ignore')
class CryptoDataImputerFixed:
"""
Specialized imputation for cryptocurrency data that preserves unique
characteristics of different crypto assets and prevents homogenization.
"""
def __init__(self, preserve_crypto_diversity=True):
self.preserve_crypto_diversity = preserve_crypto_diversity
self.crypto_profiles = {}
self.scalers = {}
def _create_crypto_profiles(self, df):
"""Create profiles for each cryptocurrency to guide imputation."""
profiles = {}
for symbol in df['symbol'].unique():
symbol_data = df[df['symbol'] == symbol]
# Calculate crypto-specific statistics
# Defensive mode extraction for 'stable' and 'blockchain_network'
stable_mode = symbol_data['stable'].mode() if 'stable' in symbol_data.columns else pd.Series()
is_stablecoin = stable_mode.iloc[0] if not stable_mode.empty else False
network_mode = symbol_data['blockchain_network'].mode() if 'blockchain_network' in symbol_data.columns else pd.Series()
blockchain_network = network_mode.iloc[0] if not network_mode.empty else None
profile = {
'symbol': symbol,
'price_level': symbol_data['price'].median() if 'price' in symbol_data.columns else None,
'price_volatility': symbol_data['price'].std() if 'price' in symbol_data.columns else None,
'volume_level': symbol_data['volume'].median() if 'volume' in symbol_data.columns else None,
'marketcap_level': symbol_data['marketcap'].median() if 'marketcap' in symbol_data.columns else None,
'dominance_level': symbol_data['dominance'].median() if 'dominance' in symbol_data.columns else None,
'rank': symbol_data['rank'].median() if 'rank' in symbol_data.columns else None,
'is_stablecoin': is_stablecoin,
'typical_rsi': symbol_data['rsi'].median() if 'rsi' in symbol_data.columns else None,
'blockchain_network': blockchain_network,
'has_onchain_data': symbol_data['transaction_count'].notna().any() if 'transaction_count' in symbol_data.columns else False,
'exchange_coverage': len([col for col in symbol_data.columns if col.startswith('symbols.') and symbol_data[col].notna().any()]),
'data_availability': len(symbol_data) / len(df) if len(df) > 0 else 0
}
profiles[symbol] = profile
return profiles
def _impute_with_crypto_context(self, df, column, crypto_profiles):
"""Impute values using crypto-specific context to prevent homogenization."""
df_result = df.copy()
for symbol in df['symbol'].unique():
symbol_mask = df['symbol'] == symbol
symbol_data = df.loc[symbol_mask, column]
if symbol_data.isnull().sum() == 0:
continue # No missing values for this symbol
profile = crypto_profiles.get(symbol, {})
is_stablecoin = profile.get('is_stablecoin', False)
rank = profile.get('rank', 999)
# Strategy depends on column type and crypto characteristics
if column in ['price', 'open', 'high', 'low', 'close']:
# Price data - special handling for stablecoins
if is_stablecoin:
# Stablecoins should stay around $1
base_price = 1.0
symbol_hash = hash(symbol + column) % 1000 / 100000 # Very small variation
adjusted_price = base_price + symbol_hash
else:
# Regular crypto - use interpolation with crypto-specific bounds
interpolated = symbol_data.interpolate(method='linear', limit_direction='both')
# If still missing, use crypto's typical price level with volatility-based noise
if interpolated.isnull().any() and profile.get('price_level'):
base_price = profile['price_level']
volatility = profile.get('price_volatility', base_price * 0.05) # Crypto is more volatile
# Add crypto-specific noise based on rank (higher rank = more volatile)
symbol_hash = hash(symbol) % 1000 / 1000 # 0-1 range
volatility_multiplier = 1 + (rank / 100) # Higher rank = higher volatility
noise_factor = (symbol_hash - 0.5) * 0.2 * volatility_multiplier # More volatile than stocks
adjusted_price = base_price * (1 + noise_factor)
else:
adjusted_price = interpolated
df_result.loc[symbol_mask, column] = symbol_data.fillna(adjusted_price)
elif column in ['volume', 'volume_alpaca']:
# Volume data - crypto volume patterns differ significantly
filled = symbol_data.fillna(method='ffill').fillna(method='bfill')
if filled.isnull().any():
base_volume = profile.get('volume_level', 1000000) # Default higher for crypto
# Major cryptos have much higher volume
if rank and rank <= 10:
volume_multiplier = 5 + (hash(symbol + column) % 1000 / 200) # 5x-10x
elif rank and rank <= 50:
volume_multiplier = 1 + (hash(symbol + column) % 1000 / 500) # 1x-3x
else:
volume_multiplier = 0.1 + (hash(symbol + column) % 1000 / 1000) # 0.1x-1.1x
adjusted_volume = base_volume * volume_multiplier
filled = filled.fillna(adjusted_volume)
df_result.loc[symbol_mask, column] = filled
elif column in ['marketcap']:
# Market cap - highly dependent on rank
if profile.get('marketcap_level'):
baseline = profile['marketcap_level']
else:
# Estimate based on rank
if rank and rank <= 10:
baseline = 10_000_000_000 # $10B+ for top 10
elif rank and rank <= 50:
baseline = 1_000_000_000 # $1B+ for top 50
elif rank and rank <= 100:
baseline = 100_000_000 # $100M+ for top 100
else:
baseline = 10_000_000 # $10M+ for others
# Add symbol-specific variation
symbol_hash = hash(symbol + column) % 1000 / 1000
baseline *= (0.5 + symbol_hash) # 0.5x to 1.5x variation
df_result.loc[symbol_mask, column] = symbol_data.fillna(baseline)
elif column in ['dominance']:
# Market dominance - only meaningful for major cryptos
if rank and rank <= 5:
# Major cryptos have meaningful dominance
symbol_hash = hash(symbol + column) % 1000 / 1000
if symbol.upper() == 'BTC':
baseline = 0.4 + (symbol_hash * 0.2) # BTC: 40-60%
elif symbol.upper() == 'ETH':
baseline = 0.15 + (symbol_hash * 0.1) # ETH: 15-25%
else:
baseline = 0.01 + (symbol_hash * 0.05) # Others: 1-6%
else:
baseline = 0.001 + (hash(symbol + column) % 1000 / 100000) # Very small
df_result.loc[symbol_mask, column] = symbol_data.fillna(baseline)
elif column in ['rsi', 'stoch_k', 'stoch_d']:
# Oscillator indicators - crypto markets are more extreme
symbol_median = symbol_data.median()
if pd.isna(symbol_median):
symbol_hash = hash(symbol + column) % 1000 / 1000
if column == 'rsi':
# Crypto RSI tends to be more extreme
if rank and rank <= 10: # Major cryptos more stable
baseline = 20 + (symbol_hash * 60) # 20-80 range
else: # Alt coins more extreme
baseline = 10 + (symbol_hash * 80) # 10-90 range
else: # stochastic
baseline = 10 + (symbol_hash * 80) # 10-90 range
else:
baseline = symbol_median
df_result.loc[symbol_mask, column] = symbol_data.fillna(baseline)
elif column in ['macd', 'macd_signal', 'macd_histogram']:
# MACD - crypto MACD values tend to be more volatile
symbol_median = symbol_data.median()
if pd.isna(symbol_median):
price_level = profile.get('price_level', 1)
symbol_hash = hash(symbol + column) % 2000 / 1000 - 1 # -1 to +1
# Scale MACD relative to price level and volatility
volatility_factor = 2 if rank and rank > 50 else 1 # Alt coins more volatile
baseline = (price_level * 0.01 * volatility_factor) * symbol_hash
else:
baseline = symbol_median
df_result.loc[symbol_mask, column] = symbol_data.fillna(baseline)
elif column.startswith('performance.'):
# Performance metrics - crypto performance is more extreme
symbol_median = symbol_data.median()
if pd.isna(symbol_median):
symbol_hash = hash(symbol + column) % 2000 / 1000 - 1 # -1 to +1
# Different baselines for different timeframes
if 'year' in column:
baseline = symbol_hash * 5 # ±500% annual performance possible
elif 'month' in column:
baseline = symbol_hash * 2 # ±200% monthly performance possible
elif 'week' in column:
baseline = symbol_hash * 0.5 # ±50% weekly performance possible
elif 'day' in column:
baseline = symbol_hash * 0.2 # ±20% daily performance possible
else: # hour, min
baseline = symbol_hash * 0.05 # ±5% short-term performance
# Alt coins are more volatile
if rank and rank > 50:
baseline *= 2
else:
baseline = symbol_median
df_result.loc[symbol_mask, column] = symbol_data.fillna(baseline)
elif column.startswith('tx_') or column.startswith('gas_') or column in [
'transaction_volume', 'transaction_count', 'total_fees', 'total_gas_used',
'avg_gas_price', 'avg_tx_size', 'fees_7d_change', 'gas_used_7d_change', 'gas_price_7d_change'
] or '_7d_change' in column:
# On-chain features - only meaningful for blockchains with transaction data
network = profile.get('blockchain_network', 'unknown')
# Special handling for 7d change columns
if '7d_change' in column:
# These are percentage changes, should be reasonable values
symbol_hash = hash(symbol + column) % 2000 / 1000 - 1 # -1 to +1 range
if 'fees' in column.lower():
# Fee changes can be more volatile in crypto
baseline = symbol_hash * 0.5 # ±50% change
elif 'gas' in column.lower():
# Gas usage changes
baseline = symbol_hash * 0.3 # ±30% change
else:
# Other transaction-related changes
baseline = symbol_hash * 0.4 # ±40% change
# Alt coins more volatile
if rank and rank > 100:
baseline *= 2
elif network in ['ethereum', 'bitcoin', 'polygon', 'bsc', 'avalanche']:
# Major networks have meaningful on-chain data
symbol_median = symbol_data.median()
if pd.isna(symbol_median):
# Estimate based on network and rank
symbol_hash = hash(symbol + column) % 1000 / 1000
if 'count' in column.lower():
if network == 'ethereum':
baseline = 1000000 * (1 + symbol_hash) # High transaction count
elif network == 'bitcoin':
baseline = 300000 * (1 + symbol_hash) # Lower transaction count
else:
baseline = 500000 * (1 + symbol_hash) # Medium transaction count
elif 'gas' in column.lower():
if network == 'ethereum':
baseline = 50 * (1 + symbol_hash) # Higher gas prices
else:
baseline = 5 * (1 + symbol_hash) # Lower gas prices
elif 'fee' in column.lower():
baseline = 1000000 * (1 + symbol_hash) # Transaction fees in wei/satoshi
else:
# Other on-chain metrics
baseline = symbol_hash * 1000
else:
baseline = symbol_median
else:
# Networks without meaningful on-chain data OR 7d_change columns
if '7d_change' in column:
# Use the calculated baseline from above
pass # baseline already set
else:
baseline = 0
df_result.loc[symbol_mask, column] = symbol_data.fillna(baseline)
elif column.startswith('exchangePrices.') or column.startswith('symbols.'):
# Exchange-specific data
exchange = column.split('.')[1] if '.' in column else 'unknown'
if column.startswith('exchangePrices.'):
# Use main price with small exchange-specific variation
main_price = profile.get('price_level', 100)
if main_price and not is_stablecoin:
# Different exchanges have small price differences
exchange_hash = hash(symbol + exchange) % 200 / 10000 # ±1% variation
baseline = main_price * (1 + exchange_hash)
else:
baseline = main_price or 1
else:
# Exchange symbols - should be strings, handle separately
continue
df_result.loc[symbol_mask, column] = symbol_data.fillna(baseline)
else:
# Generic numeric imputation with crypto-specific variation
symbol_median = symbol_data.median()
if pd.isna(symbol_median):
overall_median = df[column].median()
if pd.isna(overall_median):
overall_median = 0
# Add crypto-specific variation based on rank and volatility
symbol_hash = hash(symbol + column) % 2000 / 1000 - 1 # -1 to +1
volatility_factor = 2 if rank and rank > 100 else 1
variation = overall_median * 0.2 * symbol_hash * volatility_factor
baseline = overall_median + variation
else:
baseline = symbol_median
df_result.loc[symbol_mask, column] = symbol_data.fillna(baseline)
return df_result[column]
def _force_fill_stubborn_nulls(self, df):
"""Aggressively fill any remaining nulls with appropriate defaults."""
# Target ALL the problematic 7d_change columns
stubborn_cols = ['fees_7d_change', 'gas_used_7d_change', 'gas_price_7d_change']
for col in stubborn_cols:
if col in df.columns:
null_count_before = df[col].isnull().sum()
if null_count_before > 0:
# Strategy 1: Try group-based fill first
df[col] = df.groupby('symbol')[col].transform(lambda x: x.fillna(x.median()))
# Strategy 2: Fill remaining with symbol-specific hash-based values
still_null = df[col].isnull()
if still_null.any():
for symbol in df[still_null]['symbol'].unique():
symbol_mask = (df['symbol'] == symbol) & df[col].isnull()
if symbol_mask.any():
# Create deterministic but varied values based on symbol
symbol_hash = hash(symbol + col) % 2000 / 1000 - 1 # -1 to +1
if 'fees' in col.lower():
fill_value = symbol_hash * 0.3 # ±30% fee change
elif 'gas' in col.lower():
fill_value = symbol_hash * 0.25 # ±25% gas change
else:
fill_value = symbol_hash * 0.2 # ±20% generic change
df.loc[symbol_mask, col] = fill_value
# Strategy 3: Nuclear option - fill any remaining with 0
remaining_nulls = df[col].isnull().sum()
if remaining_nulls > 0:
print(f"[WARNING] Nuclear fill: {remaining_nulls} nulls in {col} filled with 0")
df[col] = df[col].fillna(0)
return df
def _nuclear_null_elimination(self, df):
"""Final pass to eliminate ALL nulls with extreme prejudice."""
print("[INFO] Performing nuclear null elimination...")
# Get all numeric columns
numeric_cols = df.select_dtypes(include=[np.number]).columns
for col in numeric_cols:
null_count = df[col].isnull().sum()
if null_count > 0:
print(f"[NUCLEAR] Eliminating {null_count} nulls in {col}")
# Try different strategies in order
if '7d_change' in col or 'change' in col.lower():
# Change columns - use symbol-specific hash
for symbol in df['symbol'].unique():
symbol_mask = (df['symbol'] == symbol) & df[col].isnull()
if symbol_mask.any():
symbol_hash = hash(symbol + col) % 2000 / 1000 - 1 # -1 to +1
if 'fees' in col.lower():
fill_value = symbol_hash * 0.3
elif 'gas' in col.lower():
fill_value = symbol_hash * 0.25
else:
fill_value = symbol_hash * 0.2
df.loc[symbol_mask, col] = fill_value
elif 'timestamp' in col.lower():
# Timestamp columns
df[col] = df[col].fillna(method='ffill').fillna(method='bfill').fillna(0)
elif col in ['price', 'open', 'high', 'low', 'close']:
# Price columns - use symbol-specific values
for symbol in df['symbol'].unique():
symbol_mask = (df['symbol'] == symbol) & df[col].isnull()
if symbol_mask.any():
symbol_price = df[df['symbol'] == symbol][col].median()
if pd.isna(symbol_price):
symbol_hash = hash(symbol + col) % 10000 / 100 # 0-100 range
symbol_price = 1 + symbol_hash # $1-$101
df.loc[symbol_mask, col] = symbol_price
else:
# Generic columns - try median first, then 0
median_val = df[col].median()
if pd.isna(median_val):
median_val = 0
df[col] = df[col].fillna(median_val)
# Final check - if still nulls, force to 0
remaining_nulls = df[col].isnull().sum()
if remaining_nulls > 0:
print(f"[NUCLEAR] Force filling {remaining_nulls} remaining nulls in {col} with 0")
df[col] = df[col].fillna(0)
return df
def _enhanced_sentiment_imputation(self, df):
"""Enhanced sentiment imputation that creates realistic, diverse sentiment values."""
print(f"[INFO] Starting enhanced sentiment imputation...")
# Define sentiment columns
core_sentiment_cols = ['sentiment_score', 'neg', 'neu', 'pos']
for col in core_sentiment_cols:
if col in df.columns:
null_count_before = df[col].isnull().sum()
if null_count_before > 0:
print(f"[INFO] Processing {col}: {null_count_before} nulls to fill")
# Process each symbol separately for core sentiment columns
for col in core_sentiment_cols:
if col in df.columns and df[col].isnull().any():
print(f"Enhanced imputation for {col}...")
for symbol in df['symbol'].unique():
symbol_mask = df['symbol'] == symbol
symbol_sentiment = df.loc[symbol_mask, col]
if symbol_sentiment.isnull().any():
# Try forward/backward fill first
filled = symbol_sentiment.fillna(method='ffill').fillna(method='bfill')
# For remaining nulls, use symbol-specific realistic values
if filled.isnull().any():
symbol_hash = hash(symbol + col) % 10000 / 10000
symbol_upper = symbol.upper()
# Define crypto categories
stablecoins = ['USDT', 'USDC', 'BUSD', 'DAI', 'TUSD', 'USDP']
major_cryptos = ['BTC', 'ETH', 'BNB', 'ADA', 'XRP', 'SOL', 'DOT', 'AVAX']
if col == 'sentiment_score':
# Sentiment score (-1 to +1)
if any(stable in symbol_upper for stable in stablecoins):
fill_value = (symbol_hash - 0.5) * 0.1 # Stable: ±0.05
elif any(major in symbol_upper for major in major_cryptos):
fill_value = 0.1 + (symbol_hash - 0.5) * 0.4 # Major: 0.1 ± 0.2
else:
fill_value = (symbol_hash - 0.5) * 0.6 # Alt: ±0.3
fill_value = np.clip(fill_value, -1.0, 1.0)
elif col == 'neu':
# Neutral sentiment (dominant)
if any(stable in symbol_upper for stable in stablecoins):
fill_value = 0.85 + symbol_hash * 0.1 # 0.85-0.95
elif any(major in symbol_upper for major in major_cryptos):
fill_value = 0.65 + symbol_hash * 0.2 # 0.65-0.85
else:
fill_value = 0.55 + symbol_hash * 0.3 # 0.55-0.85
fill_value = np.clip(fill_value, 0.0, 1.0)
elif col == 'pos':
# Positive sentiment
if any(stable in symbol_upper for stable in stablecoins):
fill_value = 0.05 + symbol_hash * 0.05 # 0.05-0.10
elif any(major in symbol_upper for major in major_cryptos):
fill_value = 0.15 + symbol_hash * 0.15 # 0.15-0.30
else:
fill_value = 0.10 + symbol_hash * 0.25 # 0.10-0.35
fill_value = np.clip(fill_value, 0.0, 1.0)
elif col == 'neg':
# Negative sentiment
if any(stable in symbol_upper for stable in stablecoins):
fill_value = 0.05 + symbol_hash * 0.05 # 0.05-0.10
elif any(major in symbol_upper for major in major_cryptos):
fill_value = 0.10 + symbol_hash * 0.10 # 0.10-0.20
else:
fill_value = 0.15 + symbol_hash * 0.15 # 0.15-0.30
fill_value = np.clip(fill_value, 0.0, 1.0)
filled = filled.fillna(fill_value)
df.loc[symbol_mask, col] = filled
# Normalize sentiment scores so neg + neu + pos = 1.0
if all(col in df.columns for col in ['neg', 'neu', 'pos']):
print("Normalizing sentiment scores...")
for idx in df.index:
neg_val = df.at[idx, 'neg']
neu_val = df.at[idx, 'neu']
pos_val = df.at[idx, 'pos']
current_sum = neg_val + neu_val + pos_val
if current_sum > 0:
df.at[idx, 'neg'] = neg_val / current_sum
df.at[idx, 'neu'] = neu_val / current_sum
df.at[idx, 'pos'] = pos_val / current_sum
else:
# Default neutral sentiment
df.at[idx, 'neg'] = 0.1
df.at[idx, 'neu'] = 0.8
df.at[idx, 'pos'] = 0.1
# Handle other sentiment features
other_sentiment_features = [
'social_sentiment_mean', 'social_sentiment_std', 'social_sentiment_count',
'social_confidence_mean', 'combined_sentiment', 'sentiment_agreement',
'sentiment_change_1', 'sentiment_sma_7', 'sentiment_momentum'
]
for col in other_sentiment_features:
if col in df.columns and df[col].isnull().any():
if 'sentiment' in col.lower() and 'count' not in col.lower():
# Sentiment scores - neutral with crypto-specific variation
for symbol in df['symbol'].unique():
mask = df['symbol'] == symbol
symbol_hash = (hash(symbol + col) % 200 / 1000) - 0.1 # -0.1 to +0.1
df.loc[mask, col] = df.loc[mask, col].fillna(symbol_hash)
elif 'count' in col.lower():
df[col] = df[col].fillna(0)
else:
median_val = df[col].median()
if pd.isna(median_val):
median_val = 0
df[col] = df[col].fillna(median_val)
# Final validation
print(f"[INFO] Enhanced sentiment imputation completed:")
for col in core_sentiment_cols:
if col in df.columns:
null_count_after = df[col].isnull().sum()
print(f" {col}: {null_count_after} nulls remaining")
return df
def fit_transform(self, df):
"""Apply crypto-specific imputation with anti-homogenization measures."""
df_imputed = df.copy()
df_imputed = df_imputed.sort_values(['symbol', 'interval_timestamp'])
# Create crypto profiles
self.crypto_profiles = self._create_crypto_profiles(df_imputed)
print(f"Created profiles for {len(self.crypto_profiles)} unique cryptocurrencies")
# 1. Handle categorical/flag columns
categorical_cols = [
'symbol', 'cg_id', 'blockchain_network', 'stable', 'is_crypto', 'is_stock',
'is_other', 'alpaca_data_available', 'is_trading_hours', 'is_weekend'
]
for col in categorical_cols:
if col in df_imputed.columns:
if col in ['is_crypto']:
df_imputed[col] = df_imputed[col].fillna(1) # Default to crypto
elif col in ['is_stock', 'is_other']:
df_imputed[col] = df_imputed[col].fillna(0) # Not stock/other
elif col in ['stable']:
# Determine if stablecoin based on symbol
stablecoin_symbols = ['USDT', 'USDC', 'BUSD', 'DAI', 'TUSD', 'USDP']
for symbol in stablecoin_symbols:
mask = df_imputed['symbol'].str.contains(symbol, case=False, na=False)
df_imputed.loc[mask, col] = df_imputed.loc[mask, col].fillna(True)
df_imputed[col] = df_imputed[col].fillna(False)
else:
df_imputed[col] = df_imputed.groupby('symbol')[col].fillna(method='ffill').fillna(method='bfill')
# 2. Exchange symbols (string data)
exchange_symbol_cols = [col for col in df_imputed.columns if col.startswith('symbols.')]
for col in exchange_symbol_cols:
if df_imputed[col].dtype == 'object':
# Forward/backward fill within symbol groups
df_imputed[col] = df_imputed.groupby('symbol')[col].fillna(method='ffill').fillna(method='bfill')
# 3. Core crypto market data
core_market_cols = [
'price', 'marketcap', 'volume', 'dominance', 'rank',
'open', 'high', 'low', 'close'
]
for col in core_market_cols:
if col in df_imputed.columns and df_imputed[col].isnull().any():
print(f"Imputing {col} with crypto-specific context...")
df_imputed[col] = self._impute_with_crypto_context(
df_imputed, col, self.crypto_profiles
)
# 4. Exchange prices
exchange_price_cols = [col for col in df_imputed.columns if col.startswith('exchangePrices.')]
for col in exchange_price_cols:
if df_imputed[col].isnull().any():
print(f"Imputing {col} with crypto-specific context...")
df_imputed[col] = self._impute_with_crypto_context(
df_imputed, col, self.crypto_profiles
)
# 5. Performance metrics
performance_cols = [col for col in df_imputed.columns if col.startswith('performance.') or col.startswith('rankDiffs.')]
for col in performance_cols:
if df_imputed[col].isnull().any():
print(f"Imputing {col} with crypto-specific context...")
df_imputed[col] = self._impute_with_crypto_context(
df_imputed, col, self.crypto_profiles
)
# 6. Technical indicators
tech_indicators = [
'rsi', 'macd', 'macd_signal', 'macd_histogram', 'atr', 'bb_position',
'stoch_k', 'stoch_d', 'cci', 'roc_5', 'roc_10', 'mfi', 'rsi_macd_signal',
'ema_convergence', 'true_range_pct'
]
for col in tech_indicators:
if col in df_imputed.columns and df_imputed[col].isnull().any():
print(f"Imputing {col} with crypto-specific context...")
df_imputed[col] = self._impute_with_crypto_context(
df_imputed, col, self.crypto_profiles
)
# 7. Price/volume change features
change_features = [
'price_change_1', 'price_change_7', 'price_change_14', 'volume_ratio',
'volatility_7', 'price_volume_trend', 'volatility_consistency'
]
for col in change_features:
if col in df_imputed.columns and df_imputed[col].isnull().any():
df_imputed[col] = self._impute_with_crypto_context(
df_imputed, col, self.crypto_profiles
)
# 8. On-chain features (crypto-specific) - PRIORITY HANDLING for problematic columns
onchain_features = [
'transaction_volume', 'total_fees', 'total_gas_used', 'avg_gas_price',
'transaction_count', 'tx_count_7d_change', 'tx_count_sma_7',
'tx_volume_7d_change', 'tx_volume_sma_7', 'gas_used_7d_change',
'gas_used_sma_7', 'gas_price_7d_change', 'gas_price_sma_7',
'fees_7d_change', 'avg_tx_size', 'tx_price_correlation'
]
for col in onchain_features:
if col in df_imputed.columns and df_imputed[col].isnull().any():
print(f"Imputing {col} with crypto on-chain context...")
df_imputed[col] = self._impute_with_crypto_context(
df_imputed, col, self.crypto_profiles
)
# 9. AGGRESSIVE NULL ELIMINATION for stubborn columns
df_imputed = self._force_fill_stubborn_nulls(df_imputed)
# 10. Sentiment features
sentiment_features = [
'social_sentiment_mean', 'social_sentiment_std', 'social_sentiment_count',
'social_confidence_mean', 'combined_sentiment', 'sentiment_agreement',
'sentiment_change_1', 'sentiment_sma_7', 'sentiment_momentum',
'sentiment_score', 'neg', 'neu', 'pos'
]
for col in sentiment_features:
if col in df_imputed.columns and df_imputed[col].isnull().any():
if 'sentiment' in col.lower() and 'count' not in col.lower():
# Sentiment scores - neutral with crypto-specific variation
for symbol in df_imputed['symbol'].unique():
mask = df_imputed['symbol'] == symbol
symbol_hash = (hash(symbol + col) % 200 / 1000) - 0.1 # -0.1 to +0.1
df_imputed.loc[mask, col] = df_imputed.loc[mask, col].fillna(symbol_hash)
elif 'count' in col.lower():
df_imputed[col] = df_imputed[col].fillna(0)
else:
median_val = df_imputed[col].median()
df_imputed[col] = df_imputed[col].fillna(median_val)
# 11. Quality metrics
quality_features = [
'data_quality_score', 'core_features_completeness', 'technical_indicators_completeness',
'onchain_features_completeness', 'price_data_completeness',
'overall_feature_completeness', 'data_completeness_score'
]
for col in quality_features:
if col in df_imputed.columns and df_imputed[col].isnull().any():
median_val = np.clip(df_imputed[col].median(), 0, 1)
# Add tiny crypto-specific variation
for symbol in df_imputed['symbol'].unique():
mask = df_imputed['symbol'] == symbol
symbol_hash = hash(symbol + col) % 100 / 10000 # Very small variation
fill_val = np.clip(median_val + symbol_hash, 0, 1)
df_imputed.loc[mask, col] = df_imputed.loc[mask, col].fillna(fill_val)
# 12. Temporal features
temporal_features = ['hour', 'day_of_week', 'is_weekend', 'is_trading_hours']
for col in temporal_features:
if col in df_imputed.columns and df_imputed[col].isnull().any():
if col == 'hour':
df_imputed[col] = df_imputed[col].fillna(12) # Default to noon
elif col == 'day_of_week':
df_imputed[col] = df_imputed[col].fillna(3) # Default to Wednesday
elif col == 'is_weekend':
df_imputed[col] = df_imputed[col].fillna(0) # Default to weekday
elif col == 'is_trading_hours':
df_imputed[col] = df_imputed[col].fillna(1) # Crypto trades 24/7
# 13. Handle any remaining numeric columns
remaining_numeric = df_imputed.select_dtypes(include=[np.number]).columns
remaining_with_nulls = [col for col in remaining_numeric if df_imputed[col].isnull().any()]
for col in remaining_with_nulls:
if col not in ['id', 'id_alpaca', 'backup_id'] and not col.endswith('_timestamp'):
print(f"Imputing remaining column {col}...")
df_imputed[col] = self._impute_with_crypto_context(
df_imputed, col, self.crypto_profiles
)
# 14. NUCLEAR NULL ELIMINATION - Final pass
df_imputed = self._nuclear_null_elimination(df_imputed)
print("[INFO] Crypto imputation complete with anti-homogenization measures")
return df_imputed
# Usage function with validation - FIXED VERSION
def impute_crypto_with_validation_fixed(file_path, output_path=None):
"""Impute crypto data and validate no homogenization occurred."""
try:
df = pd.read_parquet(file_path)
except Exception as e:
print(f"[ERROR] Failed to load file: {e}")
return None
# Sample symbols for validation
symbols_sample = df['symbol'].unique()[:5]
imputer = CryptoDataImputerFixed()
df_imputed = imputer.fit_transform(df)
# TRIPLE CHECK: Ensure problematic columns have no nulls
problematic_cols = ['gas_used_7d_change', 'fees_7d_change', 'gas_price_7d_change']
for col in problematic_cols:
if col in df_imputed.columns:
null_count = df_imputed[col].isnull().sum()
if null_count > 0:
print(f"[EMERGENCY] Still {null_count} nulls in {col} - applying emergency fix")
# Emergency symbol-specific fill
for symbol in df_imputed['symbol'].unique():
symbol_mask = (df_imputed['symbol'] == symbol) & df_imputed[col].isnull()
if symbol_mask.any():
symbol_hash = hash(symbol + col) % 2000 / 1000 - 1 # -1 to +1
if 'fees' in col.lower():
fill_value = symbol_hash * 0.3
elif 'gas' in col.lower():
fill_value = symbol_hash * 0.25
else:
fill_value = symbol_hash * 0.2
df_imputed.loc[symbol_mask, col] = fill_value
# Final nuclear option
df_imputed[col] = df_imputed[col].fillna(0)
print(f"[EMERGENCY] {col} nulls after emergency fix: {df_imputed[col].isnull().sum()}")
# Combine alpaca data with main data if available
price_cols = ['high', 'low', 'close', 'volume', 'open']
for col in price_cols:
alpaca_col = f"{col}_alpaca"
if col in df_imputed.columns and alpaca_col in df_imputed.columns:
df_imputed[col] = df_imputed[col].combine_first(df_imputed[alpaca_col])
# Drop unwanted columns before saving
drop_cols = [
'_filename', '_original_format', 'alpaca_data_available',
'ask_exchange', 'ask_exchange_alpaca', 'bid_exchange', 'bid_exchange_alpaca',
'conditions', 'conditions_alpaca', 'conditions_trade', 'conditions_trade_alpaca',
'symbol_quote', 'symbol_quote_alpaca', 'symbol_trade', 'symbol_trade_alpaca',
'tape', 'tape_alpaca', 'tape_trade', 'tape_trade_alpaca',
'id', 'id_alpaca', 'is_new_symbol', 'timestamp_dt',
'estimateCurrency', 'exchange', 'exchange_alpaca', 'exchange_company',
'finnhubIndustry', 'logo', 'ticker', 'weburl', 'latest_news_timestamp', 'volume_price_momentum',
'country', 'currency', 'ipo', 'name', 'period', 'phone', 'year', 'month', 'symbols.kraken',
'datetime', 'headline', 'blockchain_network', 'symbols.cryptocom', 'symbols.bitmart', 'symbols.kucoin', 'symbols.okx',
'symbols.coinbase','symbols.binance','symbols.mexc','symbols.bybit','symbols.bingx', 'symbols.huobi', 'symbols.bitget', 'symbols.gateio',
'interval_timestamp_dt', 'interval_timestamp_alpaca', 'interval_timestamp_trade', 'feature_timestamp', 'alpaca_merge_timestamp', 'sentiment_timestamp',
'hour', 'day_of_week', 'is_weekend', 'is_trading_hours', 'is_crypto', 'is_stock', 'is_other', 'gas_used_7d_change', 'fees_7d_change', 'gas_price_7d_change'
]
# Remove alpaca columns after combining
alpaca_cols = [col for col in df_imputed.columns if col.endswith('_alpaca')]
drop_cols.extend(alpaca_cols)
for col in drop_cols:
if col in df_imputed.columns:
df_imputed = df_imputed.drop(columns=col)
# Reorder columns: 'symbol' first, 'interval_timestamp' second, rest follow
cols = list(df_imputed.columns)
if 'symbol' in cols and 'interval_timestamp' in cols:
rest = [c for c in cols if c not in ['symbol', 'interval_timestamp']]
df_imputed = df_imputed[['symbol', 'interval_timestamp'] + rest]
# FINAL FINAL CHECK for problematic columns (after all drops/reorders)
for col in problematic_cols:
if col in df_imputed.columns:
null_count = df_imputed[col].isnull().sum()
if null_count > 0:
print(f"[FINAL CHECK] Still {null_count} nulls in {col} - final nuclear fill")
df_imputed[col] = df_imputed[col].fillna(0)
# Validation: Check that different symbols have different values
print("\n[VALIDATION] Checking for homogenization...")
for symbol in symbols_sample:
symbol_data = df_imputed[df_imputed['symbol'] == symbol]
if len(symbol_data) > 0:
price_mean = symbol_data['price'].mean() if 'price' in symbol_data.columns else 0
volume_mean = symbol_data['volume'].mean() if 'volume' in symbol_data.columns else 0
print(f" {symbol}: Price={price_mean:.2f}, Volume={volume_mean:.0f}")
# Save results
if output_path:
# Clean up data types
if 'backup_id' in df_imputed.columns:
df_imputed['backup_id'] = df_imputed['backup_id'].astype(str)
try:
df_imputed.to_parquet(output_path, compression='snappy')
print(f"[INFO] Crypto data imputed and saved to: {output_path}")
except Exception as e:
print(f"[ERROR] Failed to save file: {e}")
# Debug: print null count, dtype, and sample after saving
# for col in problematic_cols:
# if col in df_imputed.columns:
# print(f"[DEBUG] Nulls in {col} after save: {df_imputed[col].isnull().sum()}")
# print(f"[DEBUG] Dtype for {col}: {df_imputed[col].dtype}")
# print(f"[DEBUG] Sample values for {col}: {df_imputed[col].head(10).tolist()}")
return df_imputed
# Example usage - FIXED VERSION
def main():
input_file = "data/merged/features/crypto_features.parquet"
output_file = input_file
df_clean = impute_crypto_with_validation_fixed(input_file, output_file)
if df_clean is not None:
print(f"\n[SUCCESS] Crypto data processing completed!")
print(f"Final shape: {df_clean.shape}")
print(f"Null values remaining: {df_clean.isnull().sum().sum()}")
# Final verification of problematic columns
problematic_cols = ['gas_used_7d_change', 'fees_7d_change', 'gas_price_7d_change']
for col in problematic_cols:
if col in df_clean.columns:
nulls = df_clean[col].isnull().sum()
print(f"[FINAL VERIFICATION] {col}: {nulls} nulls")
else:
print("[ERROR] Failed to load or impute crypto data.")
if __name__ == "__main__":
main()
|