Metropolitano Plus: A Machine Learning-Based Mobile Application for Predicting Bus Arrival Times in the Corredor Metropolitano of Lima

Abstract

This study aimed to enhance the efficiency and reliability of Lima's Metropolitan Bus system by applying machine learning to predict bus arrival times and support data-driven operational management. T-RAPPI is a predictive model based on the Random Forest algorithm, trained with historical operational data from the Corredor Metropolitano. The model achieved high predictive accuracy (R2 = 0.9998, MAE = 0.0062 min), demonstrating its ability to reproduce real operational patterns. These predictions were integrated into the Metropolitano Plus mobile application, developed with Flutter and Firebase, which provides real-time bus arrival forecasts, station occupancy visualization, and trip evaluation features. By improving information reliability and reducing passenger waiting times, the proposed solution enhances both user experience and operational efficiency. A user validation survey based on the ISO/IEC 25010 quality standard reported satisfaction levels above 88% across all quality dimensions. Future work will focus on incorporating real-time traffic data and expanding the system to other public transport networks in Lima and similar urban contexts in Latin America.