Predictive Maintenance Model
AdaBoost classifier for predicting engine maintenance needs based on sensor readings.
Model Description
This model predicts whether a diesel truck engine requires maintenance (1) or is operating normally (0) based on 6 sensor inputs. It uses an AdaBoost ensemble with Decision Tree base estimators, optimized for maximum recall to minimize missed failures.
Architecture
- Algorithm: AdaBoost Classifier
- Base Estimator: Decision Tree (max_depth=3)
- Ensemble Size: 383 estimators
- Learning Rate: 0.261
Performance
| Metric | Value |
|---|---|
| Recall | 99.78% |
| Precision | 63.2% |
| F2 Score | 0.917 |
| ROC-AUC | 0.70 |
Threshold: 0.316 (optimized for maximum recall)
Usage
import joblib
from huggingface_hub import hf_hub_download
model_path = hf_hub_download(
repo_id="jskswamy/predictive-maintenance-model",
filename="best_model.joblib"
)
model = joblib.load(model_path)
prediction = model.predict_proba(features)[:, 1]
needs_maintenance = prediction > 0.316
Limitations
- Trained on Class 8 diesel truck data
- Requires all 6 sensor inputs
- Static prediction (no temporal patterns)
License
MIT License
- Downloads last month
- -