Short Ticketing Detection Framework Analysis Report

TL;DR

The paper addresses revenue loss from short ticketing in rail networks by proposing an unsupervised, multi-expert detection framework anchored on an A/B/C/D station classification. It combines four algorithms—Isolation Forest, Local Outlier Factor, One-Class SVM, and Mahalanobis Distance—into an adaptive ensemble that emphasizes diverse anomaly signals. The study identifies 30 high-risk stations and five short-ticketing patterns, delivering a detailed risk taxonomy and station-level insights that support targeted enforcement. The approach emphasizes scalability, interpretability, and automation for large networks, with future work focused on expanding data coverage, improving data fidelity, and incorporating time-period analyses. The practical impact lies in enabling revenue protection staff to focus on high-risk locations and enhancing the ability to recover lost revenue while maintaining scalable operations for transportation operators.

Abstract

This report presents a comprehensive analysis of an unsupervised multi-expert machine learning framework for detecting short ticketing fraud in railway systems. The study introduces an A/B/C/D station classification system that successfully identifies suspicious patterns across 30 high-risk stations. The framework employs four complementary algorithms: Isolation Forest, Local Outlier Factor, One-Class SVM, and Mahalanobis Distance. Key findings include the identification of five distinct short ticketing patterns and potential for short ticketing recovery in transportation systems.

Figures (2)

• Figure 1: Example illustration of a short ticketing issue.
• Figure 2: Figure on algorithm.