AshishTatapuzha/engine-predictive-maintenance-best-model

Engine Predictive Maintenance - Best Model

Model Summary

This repository contains the best-performing machine learning model for the engine predictive maintenance classification task.

Business Objective

The objective of this model is to classify engine condition as:

• 0 = Normal
• 1 = Faulty

This can help reduce unplanned maintenance, improve engine reliability, and support proactive service decisions.

Training Data

The model was trained using the prepared training split stored in the Hugging Face dataset repository:

• AshishTatapuzha/engine-predictive-maintenance-dataset
• Training file: prepared_data/train.csv
• Testing file: prepared_data/test.csv

Best Model

• Algorithm: Random Forest
• Primary selection metric: Test Recall (minimize missed faulty engines)

Features Used

• Engine rpm
• Lub oil pressure
• Fuel pressure
• Coolant pressure
• lub oil temp
• Coolant temp

Best Hyperparameters

{ "max_depth": null, "min_samples_leaf": 1, "n_estimators": 300 }

Test Performance

• Recall (primary): 0.6654
• Precision: 0.7317
• F1 Score: 0.6970
• ROC AUC: 0.6700
• Accuracy: 0.6353

Repository Contents

• best_model.joblib : serialized trained model
• feature_columns.json : list of input feature names
• best_model_params.json : best hyperparameters
• best_model_metrics.json : evaluation metrics
• model_comparison.csv : comparison of all tuned models
• all_model_tuning_results.csv : full tuning log across all models
• best_model_classification_report.csv : classification report
• requirements.txt : package versions for reproducibility

Notes

This model was selected after tuning and evaluating the following algorithms:

• Decision Tree
• Bagging
• Random Forest
• AdaBoost
• Gradient Boosting
• XGBoost