Register best predictive maintenance model

README.md

@@ -1,3 +1,77 @@
+---
+license: mit
+library_name: scikit-learn
+tags:
+- predictive-maintenance
+- engine-health
+- tabular-classification
+- sensor-data
+metrics:
+- accuracy
+- precision
+- recall
+- f1
+- roc_auc
+---
+
 # Engine Predictive Maintenance Model
 
-Best model: AdaBoost
+This repository contains the best trained model for classifying engine condition using sensor readings.
+
+## Business Objective
+
+Predict whether an engine is operating normally or requires maintenance, enabling proactive intervention before failure.
+
+## Best Model
+
+- Model selected: `AdaBoost`
+- Selection metric: F1-score
+- Target column: `Engine_Condition`
+
+## Features
+
+- `Engine_RPM`
+- `Lub_Oil_Pressure`
+- `Fuel_Pressure`
+- `Coolant_Pressure`
+- `Lub_Oil_Temperature`
+- `Coolant_Temperature`
+
+## Label Assumption
+
+- `0`: Normal/healthy operation
+- `1`: Maintenance/faulty condition
+
+## Test Metrics
+
+| model_name   |   accuracy |   precision |   recall |       f1 |   roc_auc |   best_cv_f1 | best_params                                                |
+|:-------------|-----------:|------------:|---------:|---------:|----------:|-------------:|:-----------------------------------------------------------|
+| AdaBoost     |   0.651139 |    0.648488 | 0.975233 | 0.778985 |  0.681114 |     0.775172 | {"model__n_estimators": 200, "model__learning_rate": 0.03} |
+
+## Artifacts
+
+- `best_engine_maintenance_model.joblib`: trained scikit-learn pipeline
+- `model_metadata.json`: feature list, target mapping, selected hyperparameters, metrics
+- `model_experiment_results.csv`: full model comparison
+- `requirements.txt`: dependencies for inference
+
+## Example Inference
+
+```python
+import joblib
+import pandas as pd
+
+model = joblib.load("best_engine_maintenance_model.joblib")
+
+sample = pd.DataFrame([{
+    "Engine_RPM": 800,
+    "Lub_Oil_Pressure": 3.2,
+    "Fuel_Pressure": 6.5,
+    "Coolant_Pressure": 2.4,
+    "Lub_Oil_Temperature": 78.0,
+    "Coolant_Temperature": 80.0
+}])
+
+prediction = model.predict(sample)[0]
+print(prediction)
+```

best_engine_maintenance_model.joblib

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9144ebe90152f4df37d64f2a1cd70b22c83d8ae19c0c9b3a901de5a35b50c3a6
-size 66229
+oid sha256:e2efd27063c329ed564058892c6acb7d14477512a0e05758898b7b0f84246105
+size 130229

model_experiment_results.csv

@@ -1,7 +1,7 @@
 model_name,accuracy,precision,recall,f1,roc_auc,best_cv_f1,best_params
-AdaBoost,0.6304069618633222,0.6304069618633222,1.0,0.7733124018838304,0.6764444808090487,0.7752141809940855,"{""model__n_estimators"": 100, ""model__learning_rate"": 0.05}"
-Gradient_Boosting,0.6647043767596621,0.684244167465644,0.8692651238327244,0.7657367668097281,0.700410676347899,0.7674745413885123,"{""model__n_estimators"": 100, ""model__max_depth"": 2, ""model__learning_rate"": 0.05}"
-XGBoost,0.6613770156130023,0.6796973518284993,0.8753552578156719,0.7652173913043478,0.6967336806340487,0.7703872294098518,"{""model__n_estimators"": 50, ""model__max_depth"": 3, ""model__learning_rate"": 0.05}"
-Random_Forest,0.6647043767596621,0.687479674796748,0.8583028826634186,0.7634525099313831,0.7002312901299341,0.7656012982489183,"{""model__n_estimators"": 100, ""model__max_depth"": 8, ""model__class_weight"": null}"
-Bagging,0.6542103916048119,0.698005698005698,0.7957775071051563,0.7436918990703851,0.6762816835987012,0.7426819826395361,"{""model__n_estimators"": 100, ""model__max_samples"": 0.8}"
-Decision_Tree,0.6403890453033018,0.687057991513437,0.7888753552578157,0.7344547344547344,0.6686565040718985,0.7440265771431916,"{""model__min_samples_leaf"": 1, ""model__max_depth"": 3}"
+AdaBoost,0.6511389813155875,0.6484881209503239,0.975233455136013,0.7789849197340685,0.6811138646989292,0.775172466662594,"{""model__n_estimators"": 200, ""model__learning_rate"": 0.03}"
+Bagging,0.6511389813155875,0.659976730657359,0.9212342671538774,0.7690221996271819,0.6480291977780852,0.7658860957610526,"{""model__n_estimators"": 150, ""model__max_samples"": 1.0, ""model__max_features"": 0.6}"
+Gradient_Boosting,0.6647043767596621,0.684244167465644,0.8692651238327244,0.7657367668097281,0.700410676347899,0.7674745413885123,"{""model__subsample"": 1.0, ""model__n_estimators"": 100, ""model__max_depth"": 2, ""model__learning_rate"": 0.05}"
+XGBoost,0.6613770156130023,0.6796973518284993,0.8753552578156719,0.7652173913043478,0.6967336806340487,0.7703872294098518,"{""model__subsample"": 1.0, ""model__n_estimators"": 50, ""model__max_depth"": 3, ""model__learning_rate"": 0.05, ""model__colsample_bytree"": 1.0}"
+Random_Forest,0.6649603276170976,0.687094682230869,0.8603329273244011,0.7640165855417342,0.700493059046745,0.7655985568463354,"{""model__n_estimators"": 300, ""model__min_samples_leaf"": 2, ""model__max_depth"": 8, ""model__class_weight"": null}"
+Decision_Tree,0.6403890453033018,0.687057991513437,0.7888753552578157,0.7344547344547344,0.6686565040718985,0.7440265771431916,"{""model__min_samples_split"": 2, ""model__min_samples_leaf"": 1, ""model__max_depth"": 3, ""model__class_weight"": null}"

model_metadata.json

@@ -1,5 +1,5 @@
 {
-  "created_at": "2026-05-16T08:40:48.723498Z",
+  "created_at": "2026-05-16T09:29:57.239818Z",
   "best_model_name": "AdaBoost",
   "target_column": "Engine_Condition",
   "feature_columns": [
@@ -10,18 +10,22 @@
     "Lub_Oil_Temperature",
     "Coolant_Temperature"
   ],
+  "label_assumption": {
+    "0": "normal_or_healthy",
+    "1": "maintenance_or_faulty"
+  },
   "selection_metric": "f1",
   "best_model_metrics": {
     "model_name": "AdaBoost",
-    "accuracy": 0.6304069618633222,
-    "precision": 0.6304069618633222,
-    "recall": 1.0,
-    "f1": 0.7733124018838304,
-    "roc_auc": 0.6764444808090487,
-    "best_cv_f1": 0.7752141809940855
+    "accuracy": 0.6511389813155875,
+    "precision": 0.6484881209503239,
+    "recall": 0.975233455136013,
+    "f1": 0.7789849197340685,
+    "roc_auc": 0.6811138646989292,
+    "best_cv_f1": 0.775172466662594
   },
   "best_params": {
-    "model__n_estimators": 100,
-    "model__learning_rate": 0.05
+    "model__n_estimators": 200,
+    "model__learning_rate": 0.03
   }
 }