Gendered Travel Mode Choice in Kuantan City: Optimization on Neural Network Model

Abstract

Machine learning technique is becoming increasingly prominent and extensively utilized in modelling travel mode choice to forecast mode of transportation made by travellers for a destination engagement. Neural Network is known as a powerful algorithm that can learn the complexity of human decision making; however, the role of activation function in Neural Network architecture need to be investigated to gather an informed finding of an appropriate activation function to reach optimal performance of Neural Network. This research aims to predict the Kuantan users’ travel mode choice based on gender using Neural Network model. The data was collected among 386 respondents of Kuantan City travellers. The performance of Neural Network is being trained with different activation function whilst all other parameters have been kept constant. Consequently, it has been observed that the model with the best predictive performance was created using Rectified Linear Unit (ReLU) for both male and female respondents, 0.731 and 0.735, respectively. Both group of respondents agreed that waiting time and region are the most crucial features that affecting their travel mode choice. In summary, the optimization of Neural Networks using the right activation function enhances their efficacy as robust predictive models for travel mode choice analysis, facilitating the improvement of current public transportation systems. This process ensures the neural network's capacity for high-performance prediction, contributing to more efficient and informed transportation service planning and decision-making.