| license: mit | |
| language: | |
| - en | |
| library_name: scikit-learn | |
| pipeline_tag: tabular-classification | |
| tags: | |
| - predictive-maintenance | |
| - classification | |
| - scikit-learn | |
| - tabular-data | |
| metrics: | |
| - accuracy | |
| - precision | |
| - recall | |
| - f1 | |
| # Predictive Maintenance Model | |
| ## Overview | |
| This repository contains the best-performing machine learning model developed for the predictive maintenance project. | |
| ## Business Problem | |
| The objective of this model is to classify whether an engine is operating normally or is likely to require maintenance based on sensor readings. | |
| ## Input Features | |
| - Engine_rpm | |
| - Lub_oil_pressure | |
| - Fuel_pressure | |
| - Coolant_pressure | |
| - lub_oil_temp | |
| - Coolant_temp | |
| ## Selected Model | |
| AdaBoost | |
| ## Evaluation Summary | |
| {'Model': 'AdaBoost', 'Best_Parameters': "{'learning_rate': 0.05, 'n_estimators': 100}", 'CV_Best_F1': 0.7752, 'Test_Accuracy': 0.6304, 'Test_Precision': 0.6304, 'Test_Recall': 1.0, 'Test_F1': 0.7733} | |
| ## Model Interpretation | |
| The selected model was identified after comparing multiple tree-based algorithms using cross-validation and test-set evaluation. | |
| ## Limitation | |
| Although the selected model achieved the highest test F1-score, its confusion matrix shows that it predicted all observations as class 1. This means the model was very strong in identifying maintenance-required cases but weak in distinguishing normal operating cases. | |