Revert to state from commit 4ed03954

README.md

@@ -18,7 +18,7 @@ This hospital-readmission-phase2-lgbm model predicts the risk of 30-day hospital
 
 **Task:** Hospital 30-Day Readmission Risk Prediction  
 **Model Type:** Gradient Boosting Machine (LightGBM)  
-**Training Date:** 2025-12-24 00:26:40  
+**Training Date:** 2025-12-15 10:55:05  
 **Environment:** kaggle (GPU)
 
 ## Performance Metrics
@@ -27,28 +27,28 @@ This hospital-readmission-phase2-lgbm model predicts the risk of 30-day hospital
 
 | Metric | Value |
 |--------|-------|
-| Mean ROC-AUC | 0.6772 ± 0.0054 |
+| Mean ROC-AUC | 0.8399 ± 0.0055 |
 
 ### Final Test Set Results
 
 #### Primary Metrics
 | Metric | Value |
 |--------|-------|
-| ROC-AUC | 0.6869 |
-| PR-AUC | 0.2330 |
-| F1 Score | 0.0000 |
+| ROC-AUC | 0.8424 |
+| PR-AUC | 0.4000 |
+| F1 Score | 0.1053 |
 
 #### Classification Metrics
 | Metric | Value |
 |--------|-------|
-| Precision | 0.0000 |
-| Recall | 0.0000 |
+| Precision | 0.6978 |
+| Recall | 0.0569 |
 
 #### Clinical Metrics
 | Metric | Value |
 |--------|-------|
-| Sensitivity (TPR) | 0.0000 |
-| Specificity (TNR) | 1.0000 |
+| Sensitivity (TPR) | 0.0569 |
+| Specificity (TNR) | 0.9969 |
 
 ## Model Visualizations
 
@@ -81,7 +81,7 @@ This hospital-readmission-phase2-lgbm model predicts the risk of 30-day hospital
 | Property | Value |
 |----------|-------|
 | Total Samples | 101,766 |
-| Features | 120 |
+| Features | 113 |
 | Development Set | 86,501 |
 | Final Test Set | 15,265 |
 
@@ -98,20 +98,20 @@ This hospital-readmission-phase2-lgbm model predicts the risk of 30-day hospital
 ```python
 {
   "n_estimators": 150,
-  "learning_rate": 0.01,
-  "num_leaves": 127,
+  "learning_rate": 0.05,
+  "num_leaves": 31,
   "max_depth": -1,
   "subsample": 0.7,
   "colsample_bytree": 0.7,
-  "reg_alpha": 0.1,
+  "reg_alpha": 0.0,
   "reg_lambda": 0.1
 }
 ```
 
 ## Training Details
 
-- **Total Training Time:** 417.65 minutes
-- **Hyperparameter Search Time:** 115.79 minutes
+- **Total Training Time:** 214.37 minutes
+- **Hyperparameter Search Time:** 128.14 minutes
 - **Cross-Validation Folds:** 5
 - **Early Stopping:** Yes
 - **Device:** GPU

cv_fold_details.json

@@ -2,24 +2,24 @@
   {
     "fold": 1,
     "metrics": {
-      "roc_auc": 0.6797546586916814,
-      "pr_auc": 0.21861477554247072,
-      "precision": 0.5,
-      "recall": 0.0005178663904712584,
-      "f1": 0.0010346611484738748,
-      "accuracy": 0.8883879544534998,
-      "balanced_accuracy": 0.5002264022908766,
-      "sensitivity": 0.0005178663904712584,
-      "specificity": 0.9999349381912818,
-      "ppv": 0.5,
-      "npv": 0.8884328573906006,
-      "fpr": 6.506180871828237e-05,
-      "fnr": 0.9994821336095288,
-      "true_positives": 1,
-      "true_negatives": 15369,
-      "false_positives": 1,
-      "false_negatives": 1930,
-      "brier_score": 0.09511314546587475
+      "roc_auc": 0.8422628166877643,
+      "pr_auc": 0.383648852076946,
+      "precision": 0.5921052631578947,
+      "recall": 0.04660797514241326,
+      "f1": 0.08641382621219396,
+      "accuracy": 0.8900063580139876,
+      "balanced_accuracy": 0.5212870715009399,
+      "sensitivity": 0.04660797514241326,
+      "specificity": 0.9959661678594665,
+      "ppv": 0.5921052631578947,
+      "npv": 0.8926468015627733,
+      "fpr": 0.004033832140533507,
+      "fnr": 0.9533920248575868,
+      "true_positives": 90,
+      "true_negatives": 15308,
+      "false_positives": 62,
+      "false_negatives": 1841,
+      "brier_score": 0.08057624482185294
     },
     "train_size": 62280,
     "val_size": 6920,
@@ -28,24 +28,24 @@
   {
     "fold": 2,
     "metrics": {
-      "roc_auc": 0.6831357600601401,
-      "pr_auc": 0.23023520394911784,
-      "precision": 0.0,
-      "recall": 0.0,
-      "f1": 0.0,
-      "accuracy": 0.8884393063583815,
-      "balanced_accuracy": 0.5,
-      "sensitivity": 0.0,
-      "specificity": 1.0,
-      "ppv": 0.0,
-      "npv": 0.8884393063583815,
-      "fpr": 0.0,
-      "fnr": 1.0,
-      "true_positives": 0,
-      "true_negatives": 15370,
-      "false_positives": 0,
-      "false_negatives": 1930,
-      "brier_score": 0.09581865068318382
+      "roc_auc": 0.8423694634254875,
+      "pr_auc": 0.39328310437895053,
+      "precision": 0.680327868852459,
+      "recall": 0.043005181347150256,
+      "f1": 0.08089668615984405,
+      "accuracy": 0.8909826589595375,
+      "balanced_accuracy": 0.5202338854035686,
+      "sensitivity": 0.043005181347150256,
+      "specificity": 0.997462589459987,
+      "ppv": 0.680327868852459,
+      "npv": 0.8924787518919548,
+      "fpr": 0.0025374105400130125,
+      "fnr": 0.9569948186528497,
+      "true_positives": 83,
+      "true_negatives": 15331,
+      "false_positives": 39,
+      "false_negatives": 1847,
+      "brier_score": 0.08036789416841797
     },
     "train_size": 62280,
     "val_size": 6921,
@@ -54,24 +54,24 @@
   {
     "fold": 3,
     "metrics": {
-      "roc_auc": 0.6797933866188424,
-      "pr_auc": 0.22431302959402516,
-      "precision": 0.0,
-      "recall": 0.0,
-      "f1": 0.0,
-      "accuracy": 0.8884393063583815,
-      "balanced_accuracy": 0.5,
-      "sensitivity": 0.0,
-      "specificity": 1.0,
-      "ppv": 0.0,
-      "npv": 0.8884393063583815,
-      "fpr": 0.0,
-      "fnr": 1.0,
-      "true_positives": 0,
-      "true_negatives": 15370,
-      "false_positives": 0,
-      "false_negatives": 1930,
-      "brier_score": 0.09476746711760746
+      "roc_auc": 0.8372258386399722,
+      "pr_auc": 0.3786832966547781,
+      "precision": 0.5703703703703704,
+      "recall": 0.03989637305699482,
+      "f1": 0.07457627118644068,
+      "accuracy": 0.8895375722543353,
+      "balanced_accuracy": 0.5180613940756672,
+      "sensitivity": 0.03989637305699482,
+      "specificity": 0.9962264150943396,
+      "ppv": 0.5703703703703704,
+      "npv": 0.8920477716283134,
+      "fpr": 0.0037735849056603774,
+      "fnr": 0.9601036269430052,
+      "true_positives": 77,
+      "true_negatives": 15312,
+      "false_positives": 58,
+      "false_negatives": 1853,
+      "brier_score": 0.08114327062149242
     },
     "train_size": 62280,
     "val_size": 6921,
@@ -80,24 +80,24 @@
   {
     "fold": 4,
     "metrics": {
-      "roc_auc": 0.6732777269705551,
-      "pr_auc": 0.21931957919867573,
-      "precision": 0.0,
-      "recall": 0.0,
-      "f1": 0.0,
-      "accuracy": 0.8883815028901734,
-      "balanced_accuracy": 0.5,
-      "sensitivity": 0.0,
-      "specificity": 1.0,
-      "ppv": 0.0,
-      "npv": 0.8883815028901734,
-      "fpr": 0.0,
-      "fnr": 1.0,
-      "true_positives": 0,
-      "true_negatives": 15369,
-      "false_positives": 0,
-      "false_negatives": 1931,
-      "brier_score": 0.09521447896896597
+      "roc_auc": 0.8318016530952919,
+      "pr_auc": 0.3557057624154197,
+      "precision": 0.51875,
+      "recall": 0.04298291040911445,
+      "f1": 0.07938785270205644,
+      "accuracy": 0.888728323699422,
+      "balanced_accuracy": 0.5189864125862997,
+      "sensitivity": 0.04298291040911445,
+      "specificity": 0.994989914763485,
+      "ppv": 0.51875,
+      "npv": 0.8921820303383897,
+      "fpr": 0.005010085236515063,
+      "fnr": 0.9570170895908856,
+      "true_positives": 83,
+      "true_negatives": 15292,
+      "false_positives": 77,
+      "false_negatives": 1848,
+      "brier_score": 0.08243304148954153
     },
     "train_size": 62280,
     "val_size": 6921,
@@ -106,24 +106,24 @@
   {
     "fold": 5,
     "metrics": {
-      "roc_auc": 0.6697887584277119,
-      "pr_auc": 0.21892425520456357,
-      "precision": 0.0,
-      "recall": 0.0,
-      "f1": 0.0,
-      "accuracy": 0.8883815028901734,
-      "balanced_accuracy": 0.5,
-      "sensitivity": 0.0,
-      "specificity": 1.0,
-      "ppv": 0.0,
-      "npv": 0.8883815028901734,
-      "fpr": 0.0,
-      "fnr": 1.0,
-      "true_positives": 0,
-      "true_negatives": 15369,
-      "false_positives": 0,
-      "false_negatives": 1931,
-      "brier_score": 0.09524810507805652
+      "roc_auc": 0.8460232669561988,
+      "pr_auc": 0.3943230038407428,
+      "precision": 0.6230769230769231,
+      "recall": 0.04194717762817193,
+      "f1": 0.07860262008733625,
+      "accuracy": 0.8902312138728323,
+      "balanced_accuracy": 0.5193794707842857,
+      "sensitivity": 0.04194717762817193,
+      "specificity": 0.9968117639403995,
+      "ppv": 0.6230769230769231,
+      "npv": 0.8922539312754805,
+      "fpr": 0.0031882360596004944,
+      "fnr": 0.9580528223718281,
+      "true_positives": 81,
+      "true_negatives": 15320,
+      "false_positives": 49,
+      "false_negatives": 1850,
+      "brier_score": 0.08000282103315677
     },
     "train_size": 62280,
     "val_size": 6921,

feature_importance.csv

@@ -1,121 +1,114 @@
 feature,importance
-discharge_disposition_id,181
-num_lab_procedures,157
-num_medications,145
-medical_specialty_target_encoded,133
-time_in_hospital,101
-number_diagnoses,99
-num_procedures,98
-age_bucket,78
-admission_source_id,61
-age_target_encoded,51
-number_inpatient,48
-payer_code_target_encoded,44
-admission_type_id,41
-diabetesMed_Yes,31
-diag_3_genitourinary,28
-payer_code_is_missing,26
-insulin_No,23
-medical_specialty_is_missing,21
-diag_1_respiratory,20
-change_No,20
-diag_3_diabetes,19
-diag_1_injury,18
-diag_1_circulatory,17
-diag_1_diabetes,17
-diag_2_neoplasms,16
-diag_2_diabetes,15
-diag_1_other,15
-diag_2_circulatory,14
-metformin_No,14
-age_senior,13
-diag_3_neoplasms,12
-A1Cresult_is_missing,11
-race_Caucasian,10
-metformin_Steady,9
-diag_2_respiratory,9
-age_adult,9
-max_glu_serum_is_missing,9
-insulin_Steady,8
-diag_1_genitourinary,7
-diag_1_musculoskeletal,7
-insulin_Up,6
-weight_is_missing,6
-A1Cresult_8,6
-age_elderly,6
-glipizide_No,6
-diag_2_genitourinary,6
-diag_2_other,5
-age_middle,5
-pioglitazone_No,5
-diag_3_other,5
-diag_3_respiratory,4
-gender_Male,4
-max_glu_serum_Norm,3
-diag_3_injury,3
+discharge_disposition_id,427
+diag_1_target_encoded,339
+number_inpatient_mean,300
+num_lab_procedures,284
+diag_2_target_encoded,248
+diag_3_target_encoded,248
+number_inpatient_sum,244
+med_diagnosis_interaction,171
+num_medications,146
+payer_code_is_missing,133
+diag_1_cat_target_encoded,127
+medical_specialty_target_encoded,116
+time_in_hospital,112
+number_inpatient_max,107
+diag_2_cat_target_encoded,99
+diag_3_cat_target_encoded,90
+number_diagnoses,89
+admission_type_id,78
+weight_is_missing,76
+admission_source_id,69
+total_visits,68
+payer_code_target_encoded,67
+los_medication_interaction,61
+age_target_encoded,59
+care_utilization_risk_score,56
+number_inpatient,54
+medical_specialty_is_missing,51
+age_bucket,50
+num_procedures,49
+race_Caucasian,31
+diabetesMed_Yes,30
+unique_diagnosis_categories,29
+insulin_Steady,22
+insulin_No,19
+glipizide_No,18
+A1Cresult_is_missing,12
+metformin_Steady,12
+race_Other,11
+insulin_Up,11
+weight_75_100),10
+gender_Male,10
+acarbose_Steady,9
+change_No,8
+medication_complexity_score,8
+race_is_missing,8
+weight_50_75),8
+pioglitazone_No,8
+glipizide_Steady,8
+A1Cresult_8,7
+race_Hispanic,7
+rosiglitazone_No,7
+metformin_No,7
+repaglinide_Up,6
+weight_100_125),5
+bmi_underweight,4
+max_glu_serum_Norm,4
+glyburide_No,4
+glimepiride_No,3
+bmi_normal,3
+pioglitazone_Steady,3
 glipizide_Up,3
-diag_3_circulatory,3
-glyburide_Steady,3
-diag_3_digestive,3
-glipizide_Steady,3
-diag_2_digestive,3
-glyburide_No,2
-glimepiride_No,2
-diag_2_injury,2
-pioglitazone_Steady,2
-weight_100_125),2
-max_glu_serum_300,2
-repaglinide_Steady,1
-nateglinide_Steady,1
+race_Asian,3
+glimepiride_Steady,3
+max_glu_serum_300,3
+repaglinide_No,3
+glimepiride_Up,2
+rosiglitazone_Steady,2
+glyburide_Steady,2
+bmi_overweight,2
+repaglinide_Steady,2
+weight_125_150),1
 A1Cresult_Norm,1
-repaglinide_No,1
-weight_50_75),1
-glimepiride_Steady,1
-diag_1_neoplasms,1
-diag_1_digestive,1
-diag_3_musculoskeletal,1
-metformin_rosiglitazone_Steady,0
-weight_150_175),0
-weight_125_150),0
-metformin_pioglitazone_Steady,0
-weight_25_50),0
+pioglitazone_Up,1
+glyburide_metformin_No,1
+nateglinide_Steady,1
+number_emergency,0
+bmi_obese,0
+tolazamide_Steady,0
+tolazamide_Up,0
+glyburide_metformin_Steady,0
+glyburide_metformin_Up,0
+chlorpropamide_No,0
+chlorpropamide_Steady,0
+chlorpropamide_Up,0
+acetohexamide_Steady,0
 metformin_Up,0
-repaglinide_Up,0
 nateglinide_No,0
 nateglinide_Up,0
-glimepiride_pioglitazone_Steady,0
-weight_75_100),0
-weight_175_200),0
-number_outpatient,0
-race_Asian,0
-chlorpropamide_Up,0
-chlorpropamide_Steady,0
-chlorpropamide_No,0
-weight_0_25),0
-age_very_young,0
-gender_Unknown_Invalid,0
-race_Hispanic,0
-acetohexamide_Steady,0
-diag_2_musculoskeletal,0
+weight_25_50),0
 glyburide_Up,0
 tolbutamide_Steady,0
-pioglitazone_Up,0
-rosiglitazone_No,0
-glimepiride_Up,0
-race_Other,0
+gender_Unknown_Invalid,0
+weight_0_25),0
+weight_150_175),0
+weight_175_200),0
 rosiglitazone_Up,0
-rosiglitazone_Steady,0
 acarbose_No,0
-acarbose_Steady,0
-miglitol_Steady,0
-miglitol_Up,0
 acarbose_Up,0
 miglitol_No,0
-tolazamide_Steady,0
+miglitol_Steady,0
+miglitol_Up,0
+number_outpatient_mean,0
+number_outpatient_max,0
+number_outpatient_sum,0
+number_emergency_mean,0
+number_emergency_max,0
+number_emergency_sum,0
 troglitazone_Steady,0
-number_emergency,0
-tolazamide_Up,0
-glyburide_metformin_No,0
-glyburide_metformin_Steady,0
-glyburide_metformin_Up,0
 glipizide_metformin_Steady,0
+glimepiride_pioglitazone_Steady,0
+metformin_rosiglitazone_Steady,0
+metformin_pioglitazone_Steady,0
+number_outpatient,0

gradient_boosting_metrics.json

@@ -1,20 +1,20 @@
 {
-  "roc_auc": 0.6868512620594891,
-  "pr_auc": 0.23298499648087997,
-  "precision": 0.0,
-  "recall": 0.0,
-  "f1": 0.0,
-  "accuracy": 0.8883720930232558,
-  "balanced_accuracy": 0.5,
-  "sensitivity": 0.0,
-  "specificity": 1.0,
-  "ppv": 0.0,
-  "npv": 0.8883720930232558,
-  "fpr": 0.0,
-  "fnr": 1.0,
-  "true_positives": 0,
-  "true_negatives": 13561,
-  "false_positives": 0,
-  "false_negatives": 1704,
-  "brier_score": 0.09820914777867153
+  "roc_auc": 0.8424324119878427,
+  "pr_auc": 0.3999922224076041,
+  "precision": 0.697841726618705,
+  "recall": 0.056924882629107984,
+  "f1": 0.10526315789473685,
+  "accuracy": 0.8919751064526695,
+  "balanced_accuracy": 0.5269138829486517,
+  "sensitivity": 0.056924882629107984,
+  "specificity": 0.9969028832681955,
+  "ppv": 0.697841726618705,
+  "npv": 0.8937590903080788,
+  "fpr": 0.003097116731804439,
+  "fnr": 0.943075117370892,
+  "true_positives": 97,
+  "true_negatives": 13519,
+  "false_positives": 42,
+  "false_negatives": 1607,
+  "brier_score": 0.0802369530229672
 }

gradient_boosting_model.joblib

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5f376d0d435ad8a6d5ea2106101ce17498a34e9d04d579ae8cf75808c1e55416
-size 210404
+oid sha256:655b9aa5d82389fcd7fd94d9e72e0fa88278cf0c2c55afa86378e4a9494028b3
+size 530372

training_summary.json

@@ -1,7 +1,7 @@
 {
   "model": "LightGBM Classifier",
   "task": "Hospital 30-Day Readmission Risk Prediction",
-  "timestamp": "2025-12-24 00:26:40",
+  "timestamp": "2025-12-15 10:55:05",
   "environment": "kaggle",
   "device": "gpu",
   "evaluation_pipeline": {
@@ -15,50 +15,50 @@
     "total_samples": 101766,
     "development_size": 86501,
     "final_test_size": 15265,
-    "n_features": 120
+    "n_features": 113
   },
   "best_params": {
     "n_estimators": 150,
-    "learning_rate": 0.01,
-    "num_leaves": 127,
+    "learning_rate": 0.05,
+    "num_leaves": 31,
     "max_depth": -1,
     "subsample": 0.7,
     "colsample_bytree": 0.7,
-    "reg_alpha": 0.1,
+    "reg_alpha": 0.0,
     "reg_lambda": 0.1
   },
   "cross_validation": {
-    "mean_roc_auc": 0.6771500581537861,
-    "std_roc_auc": 0.005449390770940229,
+    "mean_roc_auc": 0.839936607760943,
+    "std_roc_auc": 0.005520489368125089,
     "fold_scores": [
-      0.6797546586916814,
-      0.6831357600601401,
-      0.6797933866188424,
-      0.6732777269705551,
-      0.6697887584277119
+      0.8422628166877643,
+      0.8423694634254875,
+      0.8372258386399722,
+      0.8318016530952919,
+      0.8460232669561988
     ],
     "n_folds": 5
   },
   "final_test_metrics": {
-    "roc_auc": 0.6868512620594891,
-    "pr_auc": 0.23298499648087997,
-    "precision": 0.0,
-    "recall": 0.0,
-    "f1": 0.0,
-    "accuracy": 0.8883720930232558,
-    "balanced_accuracy": 0.5,
-    "sensitivity": 0.0,
-    "specificity": 1.0,
-    "ppv": 0.0,
-    "npv": 0.8883720930232558,
-    "fpr": 0.0,
-    "fnr": 1.0,
-    "true_positives": 0,
-    "true_negatives": 13561,
-    "false_positives": 0,
-    "false_negatives": 1704,
-    "brier_score": 0.09820914777867153
+    "roc_auc": 0.8424324119878427,
+    "pr_auc": 0.3999922224076041,
+    "precision": 0.697841726618705,
+    "recall": 0.056924882629107984,
+    "f1": 0.10526315789473685,
+    "accuracy": 0.8919751064526695,
+    "balanced_accuracy": 0.5269138829486517,
+    "sensitivity": 0.056924882629107984,
+    "specificity": 0.9969028832681955,
+    "ppv": 0.697841726618705,
+    "npv": 0.8937590903080788,
+    "fpr": 0.003097116731804439,
+    "fnr": 0.943075117370892,
+    "true_positives": 97,
+    "true_negatives": 13519,
+    "false_positives": 42,
+    "false_negatives": 1607,
+    "brier_score": 0.0802369530229672
   },
-  "hyperparameter_search_time_seconds": 6947.5521013736725,
-  "total_time_seconds": 25059.00235748291
+  "hyperparameter_search_time_seconds": 7688.534896850586,
+  "total_time_seconds": 12862.338946819305
 }