Predicting the Number of Passengers of MRT Jakarta Based on the Use of the QR-Code Payment Method during the Covid-19 Pandemic Using Long Short-Term Memory

Abstract

The trend of using public transportation has been rising over the last several decades. Because of increased mobility, public transportation has now become more crucial. In modern environments, public transportation is not only used to carry people and products from one location to another but has also evolved into a service company. In Jakarta, Mass Rapid Transit Jakarta (MRTJ) started to operate in late 2019. Recently, they updated their payment gateway system with QR codes. In this study, we predicted the hourly influx of passengers who used QR codes as their preferred payment method. This research applied machine learning to perform a prediction methodology, which is proposed to predict the number of passengers using time-series analysis. The dataset contained 7760 instances across different hours and days in June 2020 and was reshaped to display the total number of passengers each hour. Next, we incorporated time-series regression alongside LSTM frameworks with variations in architecture. One architecture, the 1D CNN-LSTM, yielded a promising prediction error of only one to two passengers for every hour.