🚗 Modeling Home-Based Trip Generation Using ANFIS

This repository provides a research-grade implementation of:

• Multiple Linear Regression (MLR)
• Adaptive Neuro-Fuzzy Inference System (ANFIS)

for modeling home-based trip generation as part of the classical four-step travel demand modeling framework.

The work reproduces and extends findings from the peer-reviewed publication:

Irshaid, M., & Abu-Eisheh, S. (2023).
Application of Adaptive Neuro-Fuzzy Inference System in Modelling Home-Based Trip Generation.
Ain Shams Engineering Journal, 14(11), 102523.
https://doi.org/10.1016/j.asej.2023.102523

Trip generation is traditionally modeled using linear statistical methods. However, real-world travel behavior is:

• nonlinear
• uncertain
• influenced by interacting socioeconomic factors

This repository investigates whether ANFIS can better capture behavioral complexity compared to classical regression models.

• Ordinary Least Squares (OLS)
• Stepwise feature selection
• Statistical significance testing
• Multicollinearity diagnostics

• Takagi–Sugeno fuzzy inference system
• Gaussian membership functions
• Hybrid learning:
  • Least Squares Estimation (forward pass)
  • Gradient Descent (backward pass)

Four trip production categories are modeled:

• Location: Salfit City, Palestine
• Source: Household travel survey
• Sample size:
  • 256 households (training/calibration)
  • 53 households (validation)

Socioeconomic variables include:

• Household size
• Employment status
• Education status
• Vehicle ownership
• Income level
• Household structure

Raw Survey Data
      ↓
Data Cleaning & Preprocessing
      ↓
Feature Engineering
      ↓
Train / Test Split
      ↓
MLR Model Training
ANFIS Model Training
      ↓
Prediction & Evaluation
      ↓
Model Comparison

• ANFIS significantly improves prediction for:

  • HBALL (complex aggregated behavior)
  • HBO (heterogeneous trip purpose)

• MLR performs competitively for:

  • HBW
  • HBE

• Insight:

  Nonlinear AI models are most beneficial when behavioral complexity increases.

home-based-trip-generation-anfis/
│
├── code/                 # Core implementation
│   ├── preprocessing.py
│   ├── mlr_model.py
│   ├── anfis_model.py
│   ├── train_mlr.py
│   ├── train_anfis.py
│   ├── evaluate_models.py
│   └── run_pipeline.py
│
├── data/                 # Dataset + metadata
│   ├── salfit_trip_data.csv
│   └── feature_description.md
│
├── figures/              # Paper figures & diagrams
│
├── notebooks/           # Exploratory & experimental analysis
│
├── results/             # Outputs & evaluation logs
│
├── paper/               # Published article
│
├── requirements.txt
└── README.md

git clone https://github.com/Mohammad-irshaid/home-based-trip-generation-anfis.git
cd home-based-trip-generation-anfis

python -m venv venv
source venv/bin/activate   # (Linux/Mac)
venv\Scripts\activate      # (Windows)

pip install -r requirements.txt

python code/run_pipeline.py

python code/train_mlr.py
python code/train_anfis.py

python code/evaluate_models.py

To reproduce the published results:

1. Load dataset from /data
2. Run preprocessing pipeline
3. Train MLR and ANFIS models
4. Evaluate using RMSE, MAE, R²
5. Compare outputs with paper results

• Single-city case study (Salfit)
• Limited sample size
• No cross-city validation
• Static socioeconomic features only

• Deep learning comparison (DNN, LSTM)
• GIS-based accessibility integration
• Transferability across cities
• Explainable AI (SHAP / LIME)
• Mode choice integration
• Activity-based modeling extension

If you use this work, please cite:

@article{irshaid2023anfis,
  title={Application of adaptive neuro-fuzzy inference system in modelling home-based trip generation},
  author={Irshaid, Mohammad and Abu-Eisheh, Sameer},
  journal={Ain Shams Engineering Journal},
  volume={14},
  pages={102523},
  year={2023},
  publisher={Elsevier}
}

• Department of Civil Engineering, An-Najah National University
• Field survey participants in Salfit City
• Transportation planning research community

This project is licensed under the MIT License.