Abstract: Systems and methods for channel selection and personalization based on machine learning models are described herein. The system receives a request to transmit a message to a recipient and identifies destination endpoint identifiers associated with the recipient. The system retrieves historical messaging data from a data store and extracts features comprising message volume metrics and channel-specific delivery metrics. The system associates the recipient with a cluster of recipients based on the extracted features and retrieves channel preference rankings for the cluster. The system selects a communication channel and destination endpoint identifier based on the channel preference rankings, then transmits the message via the selected channel. The system groups recipients based on messaging behavior similarities and derives channel preference rankings from historical delivery success rates and recipient engagement rates within each cluster.

Abstract: A computing device can assign possible error codes to subscore bins. The subscore bins can include a sent rate bin, a compliance rate bin, a fraud bin, a latency bin, and an engagement bin. The computing device can transmit network traffic messages and receive actual error codes associated with the network traffic messages. The actual error codes can overlap with and include some or all of the possible error codes. A subscore can be calculated for each subscore bin based on the actual error codes received. A deliverability score can be calculated based on the subscore for each subscore bin.