Current implementation measure the distribution of inter-arrival times (times in between two consecutive case arrival) assuming a continuous distribution.
Maybe it's better to implement the distance between the distribution of arrivals (instead of inter-arrival times):
Get all arrival timestamps (first timestamp of each trace).
Bin them, by default, by absolute hour (e.g. from 01/01/2023 08.00 am to 01/01/2023 09.00 am).
Measure distance between the two distributions.
In this way, the metric captures the temporal dimension in a better way. For example:
A process with 80 traces arriving in the first week, and 20 during the following 3 weeks; vs
A process with 20 case arrivals the first 3 weeks, and the other 80 during the last week.
Current implementation measure the distribution of inter-arrival times (times in between two consecutive case arrival) assuming a continuous distribution.
Maybe it's better to implement the distance between the distribution of arrivals (instead of inter-arrival times):
01/01/2023 08.00 amto01/01/2023 09.00 am).In this way, the metric captures the temporal dimension in a better way. For example: