Account prediction using machine learning

A method for account prediction using a trained ensemble model, comprising:

• determining that a support session has been initiated by a user with a plurality of accounts with an application;
• gathering context information related to the support session, wherein the context information includes:
• a support initiation time;
• and
• time-stamped actions preceding the support session;
• determining differences between times of the time-stamped actions and the support initiation time;
• providing input features to an ensemble model based on the differences between the times of the time-stamped actions and the support initiation time, wherein the ensemble model comprises a plurality of tree-based models and a ranking model;
• and
• determining, based on output from the ensemble model in response to the input features, an account of the plurality of accounts that the support sessions is likely to relate to.

A system, comprising:

• one or more processors;
• and a memory comprising instructions that, when executed by the one or more processors, cause the system to perform a method for training a machine learning model, the method comprising:
• receiving a historical support record comprising time-stamped actions, a support initiation time, and an account indication;
• determining features of the historical support record based at least on differences between times of the time-stamped actions and the support initiation time;
• determining a label for the features based on the account indication;
• and
• training an ensemble model, using training data comprising the features and the label, to determine an indication of an account in response to input features, wherein the ensemble model comprises a plurality of tree-based models and a ranking model.