Smart Commuting: Exploring Machine Learning Approaches to Understanding the Metro Rail Transit System

Abstract

The Metro Rail Transit Line 3 (MRT3) has been a mode of transportation for many commuters since its inception last 1999. Each day, more than thousands of passengers are recorded by the Department of Transportation (DOTr) to ride the MRT3 and predicting the number of passengers per day can be quite difficult. The ridership of the MRT3 varies daily due to factors such as holidays, working days, and even sudden technical issues. Commuters do not know how many other commuters will board on a certain day and this could lead to difficulty in planning a convenient journey to a destination. Currently, the DOTr relies on historical data plotted on spreadsheets and this can be quite difficult to analyze. In this research, a time series forecasting of daily ridership predicts the future ridership in a specific station on certain days is proposed. The proposed prediction method runs on Azure AutoML to train different models that can give accurate data and uses the ridership data from DOTr. The trained models used include Gradient Boosting, Extreme Random Trees, and Light GBM - these models have the best accuracy.