Telematics - ML based Anomaly detection system

Posted On February 12th 2025

Our customer, one of the leading Telematics data aggregators for commercial vehicles, wants to us detect Inter & Intra anomalies in the incoming telematics CAN feed and isolate erroneous data from clean data

Challenges Faced:

● Inter & Intra anomalies to be detected on the fly from millions of inflow data packets

● False alarms from Irregular feeds coming from Commercial Vehicles

Decision intelligence

● Implemented ML System for 27 Different Anomalies

● Isolation forest-based unsupervised classification and learning

● Random forest based on supervised classification and prediction

● Active pipeline for live anomaly detection available 24X7X365

● Reporting of real-time status to the user's dashboard for monitoring

Outcome delivered

● In a period of 1 year, 9 billion data packets have been examined after the system went live

● More than 15 million anomalies were prevented from entering the data warehouse.

● Cost of data quality reduced by > 90%