A Framework for Scalable Data Analysis and Model Aggregation for Public Bus Systems

Abstract

Urban mobility through quality public transportation is one of the major challenges for the consolidation of smart cities. Researchers developed different approaches for improving bus system reliability and information quality, including travel time prediction algorithms, network state evaluations, and bus bunching prevention strategies. The information provided by these approaches are complementary and could be aggregated for better predictions. In this work, we propose the architecture and present a prototype implementation of a framework that enables the integration of several approaches, which we call models, into scalable and efficient composite models. For instance, travel time prediction models can use estimators of bus position, network state, and bus headways to deliver more accurate and reliable predictions. We evaluate the scalability of the framework, the CPU usage of the framework components, and the predictions of the travel time models. We show that real-time predictions using this framework can be feasible in large metropolitan areas, such as Sao Paulo city.