Re:Equihax
by joncrocks on Thursday September 14, 2017 @12:25PM
(#15249545)
Not disagreeing with the rest but:
>And if the credit score algorithm was updated they would have to recalculate the "credit score" field on the entire database! This wouldn't really make sense from a technical perspective.
This obviously depends on your real-time requirements on providing a credit score. If you need to be able to return any credit score in X milliseconds, and you need to be able to do this at certain throughput, and have load that's not distributed across the day, then you might choose to pre-compute data.
There are techniques you could use to ensure you minimise re-computes and you could also pre-compute values into the future and re-calculate them when new/unexpected data came in.
It might not make sense to do this in a general-case, but placing retrieval performance constraints on a system can lead to non-obvious pre-calculation/computation solutions.
>And if the credit score algorithm was updated they would have to recalculate the "credit score" field on the entire database! This wouldn't really make sense from a technical perspective.
This obviously depends on your real-time requirements on providing a credit score. If you need to be able to return any credit score in X milliseconds, and you need to be able to do this at certain throughput, and have load that's not distributed across the day, then you might choose to pre-compute data.
There are techniques you could use to ensure you minimise re-computes and you could also pre-compute values into the future and re-calculate them when new/unexpected data came in.
It might not make sense to do this in a general-case, but placing retrieval performance constraints on a system can lead to non-obvious pre-calculation/computation solutions.