Public Transportation Operational Health Assessment Based on Multi-Source Data

Abstract

In order to solve the problem of inefficient long-term operation of urban public transport vehicles and the difficulty of finding the cause of the disease, a new analysis idea was designed using machine learning methods. This study aimed to provide a rapid, accurate, and convenient solution model and algorithm to address the drawbacks of traditional analysis tools that are incapable of handling multiple sources of public transport data. Based on a full process analysis of the bus operation status, the influencing factors and calculation methods were defined. Afterwards, the calculation results were used to construct a training set with a Random Forest regression model to obtain the weight ranking of different influencing factors. The results of the simulation validation proved that the model can use the basic data of bus operation to quickly find out the primary factors affecting the operation condition and pinpoint to the bottleneck interval. The method has high accuracy and feasibility. It can be universally applied to the analysis of regular bus scenarios to provide strong decision support for the operation level.

Machine Learning Applied to Public Transportation by Bus: A Systematic Literature Review

“…mass, drag coefficient), operational travel data (e.g., average speed, rate of acceleration), topological data of the road, and the battery (e.g., battery capacity, initial battery status), and Zhou et al (35) also used weather and driver behavior data obtained from unknown sources. The sources and types of data used in the solutions are described below.…”

“…Analyzing the causes of these bottlenecks can help in corrective actions. In this scenario, Zhou et al ( 35 ) evaluated the factors affecting operational functioning (e.g., road, cross, and station delays). They proposed a ML model to assess the impact of these factors.…”

“…Using basic operational data, Zhou et al ( 35 ) developed a model capable of analyzing the causes of operational bottlenecks in bus transportation. This model produces a ranking with the factors (e.g., roadway delays, intersection delays) that influence these bottlenecks’ occurrence.…”

“…Heghedus et al ( 40 ) used bus trips data, Jin et al ( 19 ) used historical data on the flow of passengers on buses, Nageshrao et al ( 28 ) used historical data on electricity usage and weather data, Delfau et al ( 32 ) used fraud statistics data and planned bus travel time as bus terminal location data, Luo et al ( 33 ) employed databases containing vehicle state information—such as positions, speeds, and direction—and vehicle-to-vehicle communication information. Abdelaty et al ( 27 ) used a database containing technical information about the bus (e.g., mass, drag coefficient), operational travel data (e.g., average speed, rate of acceleration), topological data of the road, and the battery (e.g., battery capacity, initial battery status), and Zhou et al ( 35 ) also used weather and driver behavior data obtained from unknown sources. The sources and types of data used in the solutions are described below.…”