Improving Bus Arrival Time Prediction Accuracy with Daily Periodic Based Transportation Data Imputation

Providing a predicted bus arrival time (BAT) allows bus users to choose a route with less wait time. The bus operation data needed to predict BAT is sometimes missing and must be imputed. However, the existing studies use simple methods such as last observation carried forward to impute missing bus operation data. On the other hand, in studies on traffic congestion prediction, it was reported that the prediction error was reduced by using an imputation method focusing on data characteristics. In this research, we aim to reduce the prediction error of BAT prediction by using an imputation method that focuses on the characteristics of bus operation data. We propose a pattern imputation based on the daily periodicity of bus operation data. We compared the proposed methods with a simple imputation method for BAT prediction. As a result of the experiment, we found that pattern imputation is the most effective method for imputation for multiple trips BAT prediction.