================================================================================ FEATURE ENGINEERING REPORT ================================================================================ DATASET SUMMARY -------------------------------------------------------------------------------- Input file: /app/sandbox/session_20251228_135128_cd3e46eedd35/workflow/data/consolidated_macro_data.csv Output file: /app/sandbox/session_20251228_135128_cd3e46eedd35/workflow/data/feature_engineered_data.csv Original dimensions: 396 rows × 19 features Final dimensions: 372 rows × 208 columns Date range: 1994-01-31 to 2024-12-31 Rows trimmed: 24 (6.1%) FEATURE COUNTS -------------------------------------------------------------------------------- Total features: 205 - Original features: 19 - YoY % change features: 9 - MoM change features: 7 - Rolling statistic features: 6 - Lagged features: 164 Target variables: 3 FEATURES BY CATEGORY -------------------------------------------------------------------------------- 1. Original Features: - T10Y2Y - T10Y3M - BAA10Y - VIXCLS - MARKET_INDEX - ICSA - UNRATE - CIVPART - INDPRO - TCU - RSAFS - PCE - PERMIT - HOUST - CSUSHPISA - M2SL - BUSLOANS - USREC - UMCSENT 2. YoY % Change Features: - BUSLOANS_yoy_pct - CSUSHPISA_yoy_pct - HOUST_yoy_pct - INDPRO_yoy_pct - M2SL_yoy_pct - MARKET_INDEX_yoy_pct - PCE_yoy_pct - PERMIT_yoy_pct - RSAFS_yoy_pct 3. MoM Change Features: - BAA10Y_mom - CIVPART_mom - T10Y2Y_mom - T10Y3M_mom - TCU_mom - UNRATE_mom - VIXCLS_mom 4. Rolling Statistics: - BAA10Y_vol3 - ICSA_ma3 - ICSA_ma6 - MARKET_INDEX_vol3 - VIXCLS_ma3 - VIXCLS_ma6 5. Lagged Features: 164 features (all original and transformed features with lags: [1, 3, 6, 12]) 6. Target Variables: - target_3m - target_6m - target_12m STATIONARITY ANALYSIS (ADF TESTS) -------------------------------------------------------------------------------- Total features tested: 205 Stationary features (p < 0.05): 129 (62.9%) Non-stationary features (p >= 0.05): 76 (37.1%) Non-Stationary Features (p-value >= 0.05): - PCE: p=1.0000, ADF=2.9623 - PCE_lag1: p=1.0000, ADF=2.8723 - PCE_lag3: p=1.0000, ADF=2.8460 - PCE_lag6: p=0.9991, ADF=2.7333 - PCE_lag12: p=0.9991, ADF=2.5668 - RSAFS: p=0.9979, ADF=1.6317 - RSAFS_lag1: p=0.9978, ADF=1.5923 - RSAFS_lag12: p=0.9976, ADF=1.5294 - RSAFS_lag3: p=0.9974, ADF=1.4683 - RSAFS_lag6: p=0.9966, ADF=1.2987 - M2SL_lag1: p=0.9949, ADF=1.0653 - M2SL_lag3: p=0.9945, ADF=1.0241 - M2SL_lag6: p=0.9941, ADF=0.9802 - M2SL: p=0.9928, ADF=0.8818 - M2SL_lag12: p=0.9906, ADF=0.7401 - CSUSHPISA: p=0.9857, ADF=0.5296 - MARKET_INDEX_lag1: p=0.9853, ADF=0.5148 - CSUSHPISA_lag1: p=0.9851, ADF=0.5077 - CSUSHPISA_lag3: p=0.9846, ADF=0.4922 - BUSLOANS_lag12: p=0.9843, ADF=0.4826 - CSUSHPISA_lag12: p=0.9835, ADF=0.4556 - BUSLOANS_lag6: p=0.9802, ADF=0.3657 - MARKET_INDEX: p=0.9801, ADF=0.3637 - BUSLOANS_lag3: p=0.9796, ADF=0.3523 - MARKET_INDEX_lag3: p=0.9793, ADF=0.3439 - BUSLOANS_lag1: p=0.9792, ADF=0.3426 - CSUSHPISA_lag6: p=0.9753, ADF=0.2580 - BUSLOANS: p=0.9725, ADF=0.2041 - MARKET_INDEX_lag6: p=0.9553, ADF=-0.0378 - MARKET_INDEX_lag12: p=0.9192, ADF=-0.3428 - CIVPART_lag12: p=0.9116, ADF=-0.3907 - CIVPART: p=0.9065, ADF=-0.4210 - CIVPART_lag1: p=0.9054, ADF=-0.4274 - CIVPART_lag6: p=0.9036, ADF=-0.4378 - CIVPART_lag3: p=0.8951, ADF=-0.4837 - PERMIT_lag12: p=0.6196, ADF=-1.3208 - PERMIT_lag6: p=0.6029, ADF=-1.3566 - PERMIT_lag1: p=0.6026, ADF=-1.3572 - PERMIT: p=0.6006, ADF=-1.3615 - PERMIT_lag3: p=0.5457, ADF=-1.4752 - HOUST_lag12: p=0.4248, ADF=-1.7123 - HOUST: p=0.4069, ADF=-1.7474 - HOUST_lag6: p=0.4044, ADF=-1.7522 - HOUST_lag3: p=0.4006, ADF=-1.7597 - HOUST_lag1: p=0.3979, ADF=-1.7652 - UMCSENT_lag3: p=0.3171, ADF=-1.9321 - UMCSENT_lag1: p=0.3063, ADF=-1.9558 - UMCSENT_lag6: p=0.2980, ADF=-1.9743 - UMCSENT: p=0.2894, ADF=-1.9936 - T10Y3M_lag12: p=0.1658, ADF=-2.3193 - UMCSENT_lag12: p=0.1480, ADF=-2.3782 - T10Y3M_lag6: p=0.1287, ADF=-2.4478 - CSUSHPISA_yoy_pct_lag12: p=0.1274, ADF=-2.4528 - T10Y3M_lag3: p=0.1215, ADF=-2.4755 - HOUST_yoy_pct_lag6: p=0.0906, ADF=-2.6118 - PERMIT_yoy_pct_lag6: p=0.0904, ADF=-2.6129 - PERMIT_yoy_pct_lag3: p=0.0881, ADF=-2.6242 - PERMIT_yoy_pct_lag12: p=0.0880, ADF=-2.6248 - HOUST_yoy_pct_lag1: p=0.0860, ADF=-2.6348 - HOUST_yoy_pct_lag3: p=0.0838, ADF=-2.6465 - INDPRO_lag12: p=0.0826, ADF=-2.6526 - HOUST_yoy_pct: p=0.0802, ADF=-2.6653 - INDPRO_lag6: p=0.0787, ADF=-2.6735 - HOUST_yoy_pct_lag12: p=0.0786, ADF=-2.6744 - PERMIT_yoy_pct_lag1: p=0.0765, ADF=-2.6861 - PERMIT_yoy_pct: p=0.0754, ADF=-2.6925 - INDPRO: p=0.0711, ADF=-2.7172 - INDPRO_lag3: p=0.0690, ADF=-2.7301 - INDPRO_lag1: p=0.0655, ADF=-2.7517 - CSUSHPISA_yoy_pct_lag3: p=0.0643, ADF=-2.7597 - VIXCLS_ma6_lag12: p=0.0600, ADF=-2.7878 - CSUSHPISA_yoy_pct_lag1: p=0.0583, ADF=-2.7996 - VIXCLS_ma6_lag6: p=0.0562, ADF=-2.8146 - CSUSHPISA_yoy_pct: p=0.0558, ADF=-2.8177 - VIXCLS_ma6_lag3: p=0.0537, ADF=-2.8328 - VIXCLS_ma6_lag1: p=0.0523, ADF=-2.8436 DATA QUALITY VERIFICATION -------------------------------------------------------------------------------- Missing values in final dataset: 0 Target variables preserved: True Temporal ordering maintained: True METHODOLOGICAL NOTES -------------------------------------------------------------------------------- 1. YoY % changes calculated as: (X_t - X_{t-12}) / X_{t-12} * 100 2. MoM changes calculated as: X_t - X_{t-1} 3. Rolling statistics use min_periods=window to ensure full window coverage 4. All lags use .shift() to prevent look-ahead bias 5. Rows with any NaN values dropped to ensure clean training data 6. Target variables correctly aligned with feature timestamps ================================================================================ END OF REPORT ================================================================================