MineFleet®: the vehicle data stream mining system for ubiquitous environments

This paper describes the MineFleet® distributed vehicle performance data stream mining system designed for commercial fleets. The MineFleet Onboard analyzes high throughput data streams onboard the vehicle, generates the analytics, and sends them to the remote server over the wireless networks. The paper describes the overall architecture of the system, business needs, and shares experience from successful largescale commercial deployments. MineFleet is probably one of the first distributed data stream mining systems that is widely deployed at the commercial level. The paper discusses an important problem in the context of the MineFleet ® application--computing and detecting changes in correlation matrices in a resource-contrained device that are typically used onboard the vehicle. The problem has immediate connection with many vehicle performance data stream analysis techniques such as principal component analysis, feature selection, and building predictive models for vehicle subsystems.