Application of Adaptive Neuro-Fuzzy Inference System in Modelling Home-Based Trip Generation
Publication Type
Original research

Home-based trip generation is usually modelled utilizing the traditional Multiple Linear Regression (MLR) approach. This approach may lead sometimes to unsatisfactory models, especially when the underlying relationships among the socioeconomic variables are complex and non-linear. This research investigates the feasibility of using the Adaptive Neuro-Fuzzy Inference System (ANFIS) approach in modelling home-based trip generation, and evaluates its performance as relative to the traditional approach. This is accomplished by developing four types of trip generation models for Salfit City, Palestine. These include the overall model for estimating the total daily household number of trips generated, as well as the daily home-based work, education, and others trip purposes. The outcome of the comparison between the two approaches is conducted using the Root Mean Squared Error (RMSE) and the Mean Absolute Error (MAE). The ANFIS approach is found to be a powerful tool for modelling the complex behaviour of models estimating the total daily household trips and the daily home-based other trips, compared with the MLR approach, resulting in more precise outcome and closer forecasts to real values. For example, using the ANFIS approach for the model of the total daily household number of trips reduces RMSE by 13.04% and MAE by 19.30%. However, using the traditional approach for modelling total daily home-based work and education trip purposes could be sufficient, where the R-squared is large enough to capture most of the variations among the trips, and the results of both approaches are closely compared. The new ANFIS approach represents a promising technique for modelling complex systems, which could be a good competitor for the traditional regression approach.

Proceedings of the TRB 101st Annual Meeting, Transportation Research Board
Transportation Research Board
Publisher Country
United States of America
Impact Factor
Publication Type
Both (Printed and Online)