The figure below provides a summary of our vision of a ReComp meta- process for making recurring, selective re-computation decisions on a collection of underlying analytics processes for both Forwards and Backwards patterns.
In both cases, the meta-process is triggered by observations of data changes in the environment (top). In the Forwards pattern, on the left, these are new versions of data used by the process. This pattern requires the ability to
- quantify the differences between two versions of a data,
- estimate the impact of those changes on a process outcomes,
- estimate the cost of re-computing a process,
- use those estimates to select past process instances that optimise the re-computation effort subject to a budget constraint, and
- re-enact the selected process instances, entirely or partially.
Previous: Forwards and Backwards ReComp