🚀 Vehicle Anomaly Detection & Visualization

📌 Overview

This project implements a Machine Learning-powered anomaly detection system for vehicle performance data. It integrates advanced data processing, anomaly detection, visualization, and a real-time dashboard to monitor anomalies in acceleration, braking force, and steering angle.

🏗 Project Structure

📦 Project Root
├── 📂 assets                   # Images & assets for documentation
├── 📂 data                     # Raw and processed vehicle data
│   ├── realistic_vehicle_data.csv
│   ├── processed_vehicle_data.csv
│   ├── ml_detected_anomalies.csv
│   ├── optimized_ml_detected_anomalies.csv
│   └── model_performance_metrics.csv
├── 📂 notebooks                # Jupyter Notebook for visualization
│   └── visualization.ipynb
├── 📂 plots                    # Saved visualizations of anomalies
├── 📂 src                      # Core Python scripts
│   ├── data_processing.py      # Cleans & preprocesses vehicle data
│   ├── data_simulation.py      # Generates synthetic vehicle data
│   ├── ml_anomaly_detection.py # ML anomaly detection pipeline
│   ├── visualization.py        # Generates anomaly plots
│   ├── dashboard.py            # Interactive dashboard with Dash & Plotly
├── 📂 tests                    # Unit tests for data processing
│   ├── test_data_processing.py
├── README.md                   # Project documentation

🔥 Key Features

✅ Data Processing – Cleans, normalizes, and prepares vehicle data for analysis.
✅ ML Anomaly Detection – Detects anomalies using Isolation Forest, One-Class SVM, and Local Outlier Factor.
✅ Visualization – Generates histograms, time-series plots, and scatter plots for anomaly insights.
✅ Interactive Dashboard – A real-time Dash-based UI that updates every 5 seconds.

📊 Machine Learning Anomaly Detection

The anomaly detection pipeline optimizes ML parameters and applies multiple models:

• Isolation Forest 🏕 (Best contamination level: 0.05)
• One-Class SVM 🎭 (Best kernel: rbf, nu: 0.01)
• Local Outlier Factor 🏠 (k=20, contamination= 0.03)

Final anomalies are determined through majority voting, ensuring more robust detection.

🖥 Interactive Dashboard

Run the interactive real-time dashboard with:

python src/dashboard.py

This dashboard:

• Displays acceleration, braking force, and steering anomalies
• Updates every 5 seconds to reflect new anomalies detected in real-time.

📊 Data Visualization

Jupyter Notebook notebooks/visualization.ipynb contains:

• Histograms of acceleration, braking force, and steering angle
• Time-series plots highlighting anomalies
• ML performance evaluation (Precision, Recall, F1-score, ROC AUC)

🛠 Setup Instructions

1️⃣ Clone the repository

git clone https://github.com/your-repo.git
cd your-repo

2️⃣ Install dependencies

pip install -r requirements.txt

3️⃣ Run anomaly detection

python src/ml_anomaly_detection.py

4️⃣ Launch visualization notebook

jupyter notebook notebooks/visualization.ipynb

📌 Results

✅ ML models detected anomalies in vehicle performance with ROC AUC scores around 0.48 - 0.50.
✅ Anomalies are visualized in plots & real-time monitoring is possible via the dashboard.
✅ Refined & optimized ML detection pipeline achieves better anomaly recognition.

💡 Future Improvements

• 🔥 Enhance ML models with deep learning (LSTMs for time-series)
• 📈 Integrate GPS & sensor fusion for spatial anomaly detection
• 🚀 Deploy on a cloud server for real-time vehicle monitoring

📜 License

This project is open-source under the MIT License.

---

🚀 This repository contains one of the most powerful and refined vehicle anomaly detection tools – combining ML, data science, and real-time visualization!

---

📩 Contact

If you have any questions, feel free to reach out! 😊
🔗 GitHub: s1upee
🔗 Email: lisakrasiuk@gmail.com