본문 바로가기
  • Home

Transformer Based Arrival Time Prediction for a Target Bus Stop Using Single Stop Information

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2026, 31(3), pp.37~48
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : December 26, 2025
  • Accepted : February 16, 2026
  • Published : March 31, 2026

Ga-On Cheon 1 Young-Sup Hwang 1

1선문대학교

Accredited

ABSTRACT

This study proposes a prediction method for estimating the arrival time at a destination stop when a passenger boards at any stop on the route. The data used in this work are DTG records collected every second, and we transformed these records into arrival times and travel times for each stop through preprocessing steps such as stop matching, order correction, and outlier removal. The proposed model is based on a Transformer Encoder and receives the arrival time, weather information, and location of the single stop where the passenger boards. The model predicts the travel time for each section between stops and computes the ETA to the destination by accumulating these predicted times. With this design, the model achieved superior prediction accuracy on routes with longer stop sequences compared to Curriculum learning based LSTM Model, while reducing training time by more than 50%. Experiments confirmed that destination arrival time can be accurately predicted using information from a single stop.

Citation status

* References for papers published after 2024 are currently being built.