Delay-time Prediction of the Route Bus Using Neural Network

Abstract

Currently, arrival time of route bus is estimated by a bus location system, which is frequently diverged from actual arrival time. In this study, a forecasting model for arrival time of the route bus is developed by using Neural Network, which considers several teaching signal such as stopping time at bus stops. As a result, in case correct forecast is defined as within 3 minutes differences between estimated and actual arrival time, hitting ratio is received as approximately 99% for the forecasting arrival time at next bus stop. Meanwhile, hitting ratio is declined up to 91% in case of within 1 minute range. However, hitting ratio is dropped up to approximately 80% (within 3 minutes) and 35% (within 1 minutes) in case forecasting the arrival time in the final bus stop when bus departs first bus stop. Although the hitting ratio is declined in the case forecasting arrival time at final bus stop when the bus departs first bus stop, it is approximately 10 points better result comparing to existing bus location system.