Geofences for shipping facilities and methods for estimating milestone events in shipping facilities

A method for estimating a position of carrier vessels, comprising:

• obtaining location data associated with an Automatic Identification System (AIS) device of a load being shipped to a destination location by a carrier vessel, wherein the location data indicates that the carrier vessel is proximate to one or more ports;
• providing the location data to a trained machine learning model, wherein the machine learning model is configured to infer port events, berth events, anchor events, and geofence boundary traversal events of the carrier vessel based on the synthetic port dataset;
• and
• receiving a prediction associated with the load, wherein the trained machine learning model is trained by a training algorithm that iteratively learns a geofence dataset comprising estimated ports generated from high-density clusters of AIS data and synthetic ports generated by aggregating overlapping inferred ports derived from sparse AIS data and adjusts one or more parameters of the trained machine learning model based on the geofence dataset,
• wherein the synthetic ports are formed by geometrically merging inferred ports using centroid and radius-based inference and a convex hull or union operation to define a unified port boundary,
• wherein inferred ports having fewer than a threshold number of AIS data points are removed and selectively linked to one of the estimated ports or synthetic ports based on spatial containment, boundary overlap, or proximity within a predetermined distance range, and
• wherein the geofence dataset is configured to reduce label noise and improve convergence and inference accuracy of the trained machine learning model for detecting the port events, berth events, anchor events, and geofence boundary traversal events.