Rail accidents are a major safety concern worldwide, with the U.S. witnessing over 40,000 incidents from 2001 to 2012, costing $45 million. About 5,200 accidents exceeded $141,500 in damages. The Federal Railroad Administration mandated reports with fixed fields and narratives to analyze these accidents. It utilizes text mining techniques to uncover accident characteristics, improving predictive accuracy for extreme accident costs. Results demonstrate significant enhancement in predictive accuracy, especially when using ensemble methods. Text mining provides unique insights,complementing fixed field analysis in understanding accident contributors.

• Federal Railroad Administration
• Rail Safety
• Text Mining Techniques