An ANFIS Model for Integrated Bus and Metro Travel Demand Prediction Using Automatic Data

This research explores the use of machine learning to predict public transportation demand in Mashhad, Iran. Predicting demand is crucial for optimizing operational plans and ensuring efficient service delivery. The complex nature of travel patterns necessitates a model that can account for both spatial (geographic) and temporal (time-based) factors. The developed model utilizes various spatial and temporal data points, offering flexibility and adaptability. The study compares four models built with different datasets. The research employs Adaptive Neuro-Fuzzy Inference Systems (ANFIS) to identify distinct travel patterns within each time period across the city's 253 traffic zones. Additionally, the random forest algorithm is used to identify and select the features that affect the demand variable. The most effective model leverages spatial data on an annual scale, resulting in highly accurate predictions (training error: 0.331, testing error: 1.095). This model allows planners to estimate public transportation demand across different traffic zones, both daily and annually, in response to potential changes in urban land use.