Study on Key Technologies of Transit Passengers Travel Pattern Mining and Applications based on Multiple Sources of Data

TL;DR

This paper presents advanced methodologies for mining transit passenger travel patterns using multi-source data, aiming to improve transit system efficiency through data rectification, trajectory inference, demand analysis, and passenger choice modeling.

Contribution

It introduces novel data rectification techniques, probabilistic models for trajectory inference, and comprehensive passenger demand and choice analysis methods, enhancing transit system optimization.

Findings

Improved data validity through rectification methods

Effective inference of passenger alighting points and activities

Identification of key factors influencing passenger travel choices

Abstract

In this research, we propose a series of methodologies to mine transit riders travel pattern and behavioral preferences, and then we use these knowledges to adjust and optimize the transit systems. Contributions are: 1) To increase the data validity: a) we propose a novel approach to rectify the time discrepancy of data between the AFC (Automated Fare Collection) systems and AVL (Automated Vehicle Location) system, our approach transforms data events into signals and applies time domain correlation the detect and rectify their relative discrepancies. b) By combining historical data and passengers ticketing time stamps, we induct and compensate missing information in AVL datasets. 2) To infer passengers alighting point, we introduce a maximum probabilistic model incorporating passengers home place to recover their complete transit trajectory from semi-complete boarding records.Then we propose an enhance activity identification algorithm which is capable of specifying passengers short-term activity from ordinary transfers. Finally, we analyze the temporal-spatial characteristic of transit ridership. 3) To discover passengers travel demands. We integrate each passengers trajectory data in multiple days and construct a Hybrid Trip Graph (HTG). We then use a depth search algorithm to derive the spatially closed transit trip chains; Finally, we use closed transit trip chains of passengers to study their travel pattern from various perspectives. Finally, we analyze urban transit corridors by aggregating the passengers critical transit chains.4) We derive eight influential factors, and then construct passengers choice models under various scenarios. Next, we validate our model using ridership re-distribute simulations. Finally, we conduct a comprehensive analysis on passengers temporal choice preference and use this information to optimize urban transit systems.