Abstract: A computer-implemented method includes: determining a target user corresponding to a payment request and a plurality of payment channels corresponding to the payment request; obtaining payment features based on the target user and the plurality of payment channels, in which the payment features include features representing historical characteristics of one or more payments made by the target user using one or more of the plurality of payment channels; determining, based on the payment features, a recommendation score corresponding to each of the plurality of payment channels in accordance with a prediction model; and generating, based on the determined recommendation scores of the plurality of payment channels, a recommendation including at least one of: a selection of at least one of the plurality of payment channels, and a manner of presentation of at least one of the plurality of payment channels.

Abstract: Implementations of the present specification disclose payment channel recommendation methods, apparatuses, and devices. The method includes the following: determining a target user, a target payment channel, and payment features in a current transaction when receiving a payment request; then determining a recommendation score of each target payment channel in the current transaction based on a predetermined mapping relationship; and then recommending the target payment channel based on the recommendation score, where the mapping relationship is used to represent a relationship between the payment channel and the recommendation score, and is predetermined through training on historical data of the payment features and historical data of payment result features, and the payment result features include result data specifying whether the payment channel is selected by a user and whether payment is successful.