Efficiency Assessment of Public Transport Vehicles Using Machine Learning and Non-parametric Models

Abstract

The transportation sector is essential for today’s global economy, as it tan-gents a wide range of issues such as mobility, urban planning, and economic development. Understanding the performance of vehicles is fundamental for the Brazilian economy since millions of passengers are carried by public transport every day, and this sector represents a significant share of the national GDP. Although in literature, there is a range of suitable approaches for efficiency analysis, the fourth industrial revolution has leveraged the way of acquiring data (e.g., via digital technologies), bringing the need for more advanced data analytics models to explore and process the data beforehand, as well as dealing with uncertainty. In this sense, this paper aims to provide a novel approach to assessing the efficiency of public transport vehicles by combining fuzzy clustering and Data Envelopment Analysis models in a real case study with data from embedded sensors in buses in Rio de Janeiro. A more robust integrated approach for evaluating operational efficiency can assist decision-makers and consumers in better comprehending the relationship between energy (fuel) consumption and bus efficiencies. This could enable the authorities and public transport management departments to develop appropriate policies and strategies and to reconstruct certain features of the inefficient routes, thereby increasing the operational efficiency of land transportation, reducing mobility costs, and even decreasing the carbon footprint.