Step 1: Separate the use cases into cohesive groupings.
Divide the use cases into cohesive groups of one or more use cases that together describe a function of the system.
Step 2: Assign the classes as resources for each subset of use cases
Identify the objects that the use cases use and place the representative classes into the package/partition with the use cases.
Step 3: Evaluate the ojbect model and the interaction diagrams to identify dependencies
Identify the dependencies between the packages/partitions by evaluating the associations between the participating classes and the interactions that take place between the associated objects.