Advanced Driving Cycle Generation Using XGBoost: A Case Study in Iğdır, Türkiye

Welcome to the repository for our paper, "Advanced Driving Cycle Generation Using XGBoost: A Case Study in Iğdır, Türkiye."

This repository contains the code and data used in our study, which presents a novel framework for driving cycle generation leveraging On-Board Diagnostics (OBD) data and the XGBoost algorithm. The case study focuses on Igdir, Turkey, known for its complex and diverse urban traffic patterns. Our framework outperforms traditional methods like Markov chains, offering more accurate and region-specific tools for vehicle performance and emissions evaluation.

---

Project Structure

The project is organized as follows:

IgdirDrivingCyclesML/
├── data/
│   ├── processed/                 # Processed datasets
│   │   ├── combined_data.csv       # Combined driving data from all cycles
│   │   ├── cycle_lengths.txt       # Length of each driving cycle
│   │   └── features_data.csv       # Feature-engineered dataset
│   └── raw/                        # Raw datasets
│       ├── D01.txt                 # Raw driving data for cycle D01
│       ├── D02.txt                 # Raw driving data for cycle D02
│       ├── ...                     # (Additional raw data files)
│       └── segment_sizes.mat       # MATLAB file containing segment sizes
├── docs/                           # Documentation
├── models/                         # Trained models
│   └── xgboost_model.pkl           # Serialized XGBoost model
├── notebooks/                      # Jupyter notebooks
│   └── driving_cycle.ipynb         # Notebook for analysis
├── references/                     # References and scripts
├── reports/                        # Reports, figures, and metrics
├── src/                            # Source code
├── README.md                       # This README file
├── environment.yml                 # Conda environment configuration
└── requirements.txt                # Python dependencies

Credits and Acknowledgments

The Jupyter notebook notebooks/driving_cycle.ipynb is adapted from https://github.com/fesihkeskin/Driving_Cycle. This adaptation was made to facilitate comparisons between our study and the methods presented in https://doi.org/10.1177/03611981241260700.

---

Installation

Prerequisites

• Python 3.7 or higher
• Conda (recommended) or Python's virtualenv

Setting Up the Environment

1. Clone the repository:

   git clone https://github.com/username/IgdirDrivingCyclesML.git
   cd IgdirDrivingCyclesML

2. Create the Conda environment:

   conda env create -f environment.yml
   conda activate igdir-driving-cycles

   Alternatively, if you're using virtualenv, install dependencies with:

   pip install -r requirements.txt

---

Usage

Step 1: Data Processing

python src/data/make_dataset.py

Step 2: Feature Engineering

python src/features/build_features.py

Step 3: Model Training

python src/models/train_model.py

Step 4: Model Prediction

python src/models/predict_model.py

Step 5: Manual Data Split and Testing

python src/models/manual_split_test.py

---

Documentation and Visualizations

• Figures: Saved in reports/figures/.
• Reports and Logs: Saved in reports/.

---

Citation

If you use this repository in your research or work, please cite our paper:

@InProceedings{Keskin2024,
  author    = {Keskin, Fesih},
  booktitle = {2024 8th International Artificial Intelligence and Data Processing Symposium (IDAP)},
  title     = {Advanced Driving Cycle Generation Using XGBoost: A Case Study in Iğdır, Türkiye},
  year      = {2024},
  month     = sep,
  pages     = {1--6},
  publisher = {IEEE},
  doi       = {10.1109/idap64064.2024.10710756},
}

F. Keskin, "Advanced Driving Cycle Generation Using XGBoost: A Case Study in Iğdır, Türkiye," 2024 8th International Artificial Intelligence and Data Processing Symposium (IDAP), Malatya, Turkiye, 2024, pp. 1-6, doi: 10.1109/IDAP64064.2024.10710756.

---

License

The code and data in this repository are licensed under the MIT License.

---

Contact

---

For questions or contributions, please open an issue or pull request!

---