northeastern university seal

Anomaly Detection Using Manifold Embedding and its Applications in Transportation Corridors

Topics: Transportation,

This article investigates manifold embedding tools to detect anomalous cargo in secure transportation corridors. Both linear methods like multi-dimensional scaling and nonlinear methods like "locally linear embedding" and isometric feature mapping are explored through experiments on simulated data and actual truck data from weigh stations. It finds that manifold embeddings in general, but even more so those that are nonlinear, are "surprisingly natural" for representing data to detect anomalies.

View Source