Read case study · Project index · Engineering guide
Archived publication title: Predicting Vehicle Type Using NGSIM US-101. The Zenodo record retains that citation title.
Project author and sole implementer: Mohammed Mahyoub.
The complete source notebook and case-study gallery provide the evaluation context and remaining figures.
This project applies supervised machine learning techniques to classify vehicle types using the NGSIM US-101 trajectory dataset.
The objective is to compare the performance of three classification algorithms:
- Logistic Regression
- Decision Tree
- Random Forest
Vehicle trajectories are aggregated to vehicle level to prevent data leakage and provide independent observations for model training and evaluation.
- Dataset: NGSIM US-101 Vehicle Trajectories
- Source: Federal Highway Administration (FHWA)
- Classification task:
- Motorcycle
- Passenger Car
- Truck
The models use the following traffic and vehicle characteristics:
- Vehicle Length
- Vehicle Width
- Mean Speed
- Mean Acceleration
- Dominant Lane
- Mean Space Headway
- Logistic Regression
- Decision Tree
- Random Forest
Model performance is evaluated using:
- Accuracy
- Precision
- Recall
- F1-score
- Confusion Matrix
- Python
- Pandas
- NumPy
- Matplotlib
- Seaborn
- Scikit-learn
- Jupyter Notebook
- Predicting_Vehicle_Type_Using_NGSIM_US101.ipynb
Mohammed Mahyoub Ali Ayedh Mohammed
MSc Artificial Intelligence
University of the West of England (UWE Bristol)
This repository is intended for educational and academic purposes.