Abstract: An intermediate result set is obtained from a first machine learning model whose input data set comprises values of a first set of properties associated with a session of operations. The intermediate result and a second set of properties associated with the session are provided as input to a second machine learning model. A value of at least one property of the first set is determined before a value of at least one property of the second set is determined. A particular action recommendation, based at least in part on output generated by the second machine learning model, is implemented.

Abstract: Described are techniques for determining a confidence value associated with the probability that a user will access a particular second user interface by interacting with a first user interface. The confidence value may be determined based on user interaction data indicative of prior interactions by the user. Based on the confidence value, the second user interface may be pre-generated, in advance of user interaction with the first user interface. If the user interacts with the first user interface in a manner that would cause presentation of the second user interface, the user may be provided with the second user interface promptly, avoiding the latency that would normally be experienced during the time that the second user interface is generated. Pre-generation of only the user interfaces indicated by the user interaction data may limit unnecessary use of computing resources.