Predictive Automotive Maintenance

In this project, I developed a predictive maintenance model using automotive engine data. The goal was to forecast potential engine failures before they occur, allowing vehicle owners and mechanics to perform maintenance and reduce unexpected breakdowns.

Using Python and machine learning libraries, I analyzed key engine parameters and trained classification models to predict failure likelihood. Through exploratory data analysis, feature engineering, and model evaluation, I identified the most influential factors in engine degradation. The final model achieved high accuracy.

This project demonstrates my ability to translate business needs into actionable solutions, combining domain knowledge with technical skills to improve operational efficiency and extend engine lifespans.

Overview:
In the world of automotive diagnostics, early detection of engine issues is crucial for performance, cost efficiency, and safety. This machine learning project focuses on predictive maintenance using real-time engine telemetry data to identify signs of wear and potential failure before they become critical.

Objective:
The goal was to build a classification model that empowers vehicle owners and mechanics to detect engine anomalies in advance, enabling proactive repairs and minimizing costly breakdowns.

Dataset Highlights:
• Total Entries: 19,535
• Features: Engine RPM, Lubrication Oil Pressure, Fuel Pressure, Coolant Pressure, Oil Temperature, Coolant Temperature
• Target Variable: engine_condition
o 0 = Normal
o 1 = Needs Maintenance

Tools & Technologies Used:
• Programming: Python
• Data Processing: Pandas, Scikit-learn
• Modeling: Random Forest, XGBoost
• Visualization: Matplotlib, Seaborn
• Notebook Interface: Jupyter
• Deployment (Optional): Streamlit

Exploratory Data Analysis (EDA):
Performed detailed EDA to identify relationships between variables and assess class distributions. Generated correlation heatmaps, feature distributions, and class imbalance metrics to guide model development.

Model Performance:
After testing multiple algorithms, a Random Forest classifier provided the best results:
• Accuracy: ~65%
• Precision (Maintenance Class): 69%
• Recall (Maintenance Class): 80%
These results show strong sensitivity to maintenance needs, which is critical in predictive systems.

Key Visuals:
• Correlation Heatmap
• Confusion Matrix
• ROC Curve
• Feature Importance Plot

How It Works:
1. Load telemetry data
2. Apply preprocessing and scaling
3. Use the trained model to classify engine health

Project Repository Includes:
• Data file (engine_data.csv)
• Notebooks for EDA and model training
• Pre-trained model and scaler
• Visual assets
• Deployable app script (app.py)

Impact:
This project bridges the gap between raw telemetry data and actionable insights. By integrating machine learning with automotive diagnostics, it delivers a tool that could enhance operational efficiency in auto repair shops or vehicle fleet monitoring systems.