Predicting willingness to pay for urban rail transit using machine learning : Evidence from jakarta MRT

Abstract

The development of urban transportation requires an efficient, reliable and sustainable system, so fare determination is an important factor in the success of the Jakarta MRT service. In this context, understanding the user's Willingness to Pay (WTP) is crucial because it is not only influenced by economic ability, but also perception and preference for services. This study aims to analyze and predict the WTP of MRT users by integrating transportation economics approaches and machine learning methods. The research data is in the form of primary data from a survey of 296 MRT users which includes socio-economic characteristics, transportation costs, frequency of use and Ability to Pay (ATP). The methodology used includes descriptive analysis and regression modeling using various algorithms, namely Linear Regression, Decision Tree, Random Forest, Gradient Boosting, Support Vector Regression (SVR) and XGBoost. Model evaluation was carried out using MAE, RMSE and determination coefficient (R²).

The results showed that the value of WTP was relatively homogeneous compared to variations in income and transportation costs, which indicated that willingness to pay was not entirely determined by economic ability. The performance of the model shows that no algorithm is consistently superior, with R² values that tend to be low. The feature importance analysis identified income, transportation costs and ATP as the main factors. This research contributes through the application of a multi-model machine learning framework and policy implications that MRT fare determination needs to consider economic aspects and user preferences in a balanced manner.