This study investigates the feasibility of using the Adaptive Neuro-Fuzzy Inference System (ANFIS) in modelling home-based trip generation, and evaluates its performance as relative to the Multiple Linear Regression (MLR). The implementation methodology and the underlying principles of ANFIS architecture are also discussed. This is accomplished by developing different trip generation models for Salfit City, Palestine. These include the overall model for estimating total daily household trips generated (HBALL), as well as the daily home-based work (HBW), education (HBE), and others (HBO) trip purposes. The outcome of the comparison between the two approaches is performed using the Root Mean Squared Error (RMSE), the Mean Absolute Error (MAE), and R-squared. The results indicate that ANFIS is a powerful tool for modelling HBALL and HBO trips, which encounter more complicated behaviour with wider data ranges and larger average number of trips per day. ANFIS showed better accuracy and closer predictions compared to the MLR approach. For example, modelling HBALL trips using ANFIS approach, with three Gaussian-type membership functions for each input variable and 80 training epochs, resulted in RMSE of 1.4880 compared with 1.7112 using MLR, with a reduction of 13.04%. On the other hand, using the MLR approach for modelling HBW and HBE trip purposes, with less complicated behaviour, might be sufficient. The R-squared using the MLR is large enough to capture most of the variation among the trips, and the results of the two approaches are closely comparable. R-squared using the MLR approach is either almost equal to that of the ANFIS approach for the HBE trips, or slightly less for the HBW trips. Overall, ANFIS approach is found to be a promising technique for modelling systems with complex behaviour. Its potential for further exploration in transportation research is recommended.

Title

Ain Shams Engineering Journal

Publisher

Elsevier

Publisher Country

Egypt

Indexing

Scopus

Impact Factor

3.76

Publication Type

Both (Printed and Online)

Volume

14

Year

2023

Pages

11