Abstract: Computer systems and associated methods are disclosed to implement a machine-learned shipping or delivery destination prediction model that can be used to predict shipping or delivery destinations for customer shipping or delivery requests. The prediction model may be trained using a machine learning technique to select, based on customer data, one or more location classes that correspond to different types of location data found in the customer data. Depending on the embodiment, the different types of location data may include a customer's most frequent shipping address, last shipping address, billing address, email address, or others. In embodiments, the prediction model may be used in an item ordering service to auto-populate the shipping or delivery address for a user interface or to provide location-specific shipping options or offers. In embodiments, performance of the prediction model may be determined based on actual customer feedback, and tracked to provide a model performance report.