Machine-learning processing at native-location storage system to generate collections action plan

Techniques are disclosed for using machine-learning processing for generating resource-allocation specifications. A first data set may be received from a first data source. The first data set can include a first resource request and a first timestamp associated with entities. A second data set can be received from a second data source that includes communication data and allocation data associated with the entities. Target characteristics may be defined for training instances. The training instances can be used to train a machine-learning model using the first data set and the second data set. A third data set may be accessed and used to generate a user session within which, the trained machine-learning model may execute to generate a resource-allocation specification. The resource-allocation specification including a communication schedule. One or more communications compliant with the communication schedule may be output to an entity.