Validation method for determining high-risk time and locations of graffiti attacks on trains

Graffiti attacks on trains pose considerable business challenges due to negative financial, logistic and social factors. Prevention of graffiti attacks is the key to solving these challenges and can be improved by targeted security-guard deployment to potential hot-spots. Identifying potential hot-spots is a problem ideally suited for a Data Science solution, but is not without methodological challenges.

This case describes our approach to gain insight into risky times and locations as well as a novel method for validating those potential hot-spots.

The first part of the talk will detail our approach for prediction of graffiti attacks which integrates multiple data sources, including GPS location of trains, weather and sun radiation data, school holidays information, and graffiti observation reports. The second part of the talk will primarily focus on the validation method employed. The validation method posed particular challenges because the vast majority of graffiti attacks occur undetected, which results in sparse security reports records.

The talk will describe our approach for validating results by utilizing simulations based on limited security reports data. In conclusion, the data-driven approach and innovative validation method described in this case have a potential to enable effective mitigation of graffiti attacks on trains, thereby enhancing proactive security measures and promoting a safer transportation environment.