Detecting Labor Data Anomalies with Outlier Voting
In 2024, an Amtrak Continuous Improvement team began analyzing Amtrak’s timekeeping practices in the infrastructure maintenance division to improve detection of non-obligated payments and timekeeping discrepancies. The team was quickly overwhelmed by vast structured and unstructured data, legacy timekeeping systems with limited controls, and timekeeping landscape governed by multiple collective bargaining agreements.
Continuous Improvement at Amtrak has traditionally focused on applying Lean and Six Sigma methodologies. While these approaches are highly effective, the team believed that advanced analytic techniques were needed to address the challenges associated with labor data. In June 2025, the team partnered with Amtrak’s Operations Research group to pilot various machine learning and advanced analytic techniques to timekeeping, including outlier detection algorithms for anomalies at the work group level. This presentation details the approach, timeline, key decision points, results, lessons learned, and future state for labor data analysis.
Professional