BART's platform signs tell riders when the next train arrives and how many cars it has. Before our project, BART had no way to know when those signs were wrong. They found out when customers complained.

We analyzed BART's train arrival and sign prediction data to find patterns in when and where signs were inaccurate, and built interactive dashboards so the control room could monitor sign accuracy in real time. Senior team project at USF, Fall 2020, sponsored by BART engineer Herbert Diamant. Collaborated with a UC Berkeley research group and Splunk's team on Azure.

BART platform sign showing train ETAs Splunk dashboard showing average hourly time deviation by day of week Splunk transaction showing correlated door open and DSS ETA events

Shoutout to my teammates Antonio Gutierrez and Summer Marsh, our sponsor Herbert Diamant at BART, our instructors Doug Halperin and Luigi Lucaccini, and our TA Aarthi Parthipan.