Predicting bus arrival times (BAT) benefits both users and operators. Previous studies assumed using bus operation data, but missing data occur frequently. Simple methods such as LOCF were used to supplement missing data. In traffic congestion prediction, it was reported that errors could be reduced by focusing on data characteristics. Thus, we focus on imputation methods for bus operation data to reduce prediction errors.
Buses have daily periodicity, allowing for expected daily run and stop time cycles. We apply imputation methods that reflect daily periodicity to actual bus operation data and analyze the changes in BAT prediction errors.
Publication
- Takumi Niwa, Ismail Arai, Arata Endo, Masatoshi Kakiuchi, Kazutoshi Fujikawa, “Improving Bus Arrival Time Prediction Accuracy with Daily Periodic Based Transportation Data Imputation,” IEEE International Conference on Smart Mobility (SM) (SM'2023), Mar. 2023.