Predictive Process Monitoring Feature Factory Not Progressing

[TKForgeron]

Hi,

I'm constructing my Feature_Storage object using ocpa.algo.predictive_monitoring.factory.apply(), but it does not load. The issue arises when I work with the order management log (running-example.jsonocel). With the custom industrial OCEL I have it works fine, and also with BPI17 it does not show any troubles.

I have tried a few things already:

• use LEAD_TYPE instead of CONN_COMP
• try different object types as leading type.
• exclude "customers" as an object type (I've transformed the OCEL to have this as event attributes)
• exclude event attributes from the Feature_Storage construction factory.

I understand that it takes longer because of the longer OC process executions, but I cannot understand why it does not run at all. Even more so, as I've run it with other datasets. Moreover, I've seen the code of your last publication (preserving OC structures), and there it seems to work.

Can you help me with this?

Thanks in advance!

Here is an excerpt from my notebook (I've also let it run 12 hours before, but no luck):

Here is the script I'm using:

import ocpa.objects.log.importer.ocel.factory as ocel_import_factory
import ocpa.algo.predictive_monitoring.factory as feature_factory
ocel = ocel_import_factory.apply(
ocel_in_file, parameters={"execution_extraction": "leading_type","leading_type": "orders"}
)
all_acts = ocel.log.log["event_activity"].unique()
event_level_features = (
{
f"C2{char}": (feature_factory.EVENT_PRECEDING_ACTIVITIES, (act,))
for act, char in zip(all_acts, string.ascii_lowercase)
}
| {
"P2": (feature_factory.EVENT_ELAPSED_TIME, ()),
"P3": (
feature_factory.EVENT_REMAINING_TIME,
(),
),
# "P4": (feature_factory.EVENT_FLOW_TIME, ()),
"P5": (feature_factory.EVENT_SYNCHRONIZATION_TIME, ()),
# "P6": (feature_factory.EVENT_SOJOURN_TIME, ()),
}
| {
f"O3{char}": (feature_factory.EVENT_PREVIOUS_TYPE_COUNT, (ot,))
for ot, char in zip(objects_metadata, string.ascii_lowercase)
}
)
feature_storage = feature_factory.apply(
ocel,
event_based_features=list(event_level_features.values()),
event_attributes=[f"event_{ea}" for ea in event_attributes]
) # takes much too long when 'customers' is an object type