BPIC17 Evaluation

BPIC17 - Evaluation

This page describes how to reproduce the BPIC17 experiment using the OCEAN library. The resulting aggregated EKG is consistent with the visualization reported in: Multi-Dimensional Event Data in Graph Databases

---

Applied Steps

AggrStep(aggr_type="ENTITIES", ent_type="Application", group_by=[ekg.type_tag], where=None, attr_aggrs=[]),
AggrStep(aggr_type="ENTITIES", ent_type="Offer", group_by=[ekg.type_tag], where=None, attr_aggrs=[]),
AggrStep(aggr_type="ENTITIES", ent_type="Workflow", group_by=[ekg.type_tag], where=None, attr_aggrs=[]),
AggrStep(aggr_type="ENTITIES", ent_type="Resource", group_by=[ekg.type_tag], where=None, attr_aggrs=[]),
AggrStep(aggr_type="EVENTS", ent_type= None, group_by=[log.event_activity,'lifecycle'], where=None, attr_aggrs=[])]

Reproduce the Evaluation

Import the BPIC17 EKG

Follow the PromG tutorial to load the BPIC17 EKG: https://github.com/PromG-dev/ekg_bpic17/tree/main

Create a working directory with the following files

• ekg_config.yaml:

type_tag: "Type"    # how the Type of an entity, df, class, etc. is called

entity_type_mode: 'label'  # specify if the entity type is represented as a label or as a property in the graph database  
# label = :Entity:EntityType
# property = :Entity (Type: 'EntityType')

# Neo4j configuration
neo4j:
  URI: '<YOUR-URL>'
  username: '<YOUR-USERNAME>'
  password: '<YOUR-PASSWORD>'

• log_config.yaml (this already corresponds to the structure of the EKG exported with PromG):

event_id: "ID"    # unique identifier attribute for the event
event_activity: "activity"   # activity attribute name of the event
event_timestamp: "timestamp"    # timestamp attribute name of the event

entity_id: "sysId"   # unique identifier attribute for the entity

• main.py:

from ocean_lib import pipeline, AggrSpecification, AggrStep

@pipeline(first_load=False)
def build_aggr_spec(log, ekg):
    aggr_basic = [
        AggrStep(aggr_type="ENTITIES", ent_type="Application", group_by=[ekg.type_tag], where=None, attr_aggrs=[]),
        AggrStep(aggr_type="ENTITIES", ent_type="Offer", group_by=[ekg.type_tag], where=None, attr_aggrs=[]),
        AggrStep(aggr_type="ENTITIES", ent_type="Workflow", group_by=[ekg.type_tag], where=None, attr_aggrs=[]),
        AggrStep(aggr_type="ENTITIES", ent_type="Resource", group_by=[ekg.type_tag], where=None, attr_aggrs=[]),
        AggrStep(aggr_type="EVENTS", ent_type= None, group_by=[log.event_activity,'lifecycle'], where=None, attr_aggrs=[])]
    
    return AggrSpecification(aggr_basic)

See Performance

Step	#Aggregated Nodes	Query Exec Time (s)	#nodes / ms
Aggr Entity	31509	0,552	51,1
Aggr Entity	42995	0,297	144,68
Aggr Entity	31509	0,209	150,14
Aggr Entity	149	0,01	14,73
Aggr Event	1202267	16,807	71,53
Finalization	117499	1,55	75,81
Rel Inference	-	13,115	-

Open Visualization

1. Open Neo4j Bloom.
2. Go to Perspectives > Saved Cypher.
3. Create a new Search Phrase and name it, for example, BPIC17.
4. In the Cypher query field insert:

MATCH (c1:Class)-[df:DF_C]->(c2:Class)
WHERE df.Type <> 'Resource'
  AND df.count >= 500

AND NOT (
    df.Type STARTS WITH 'CASE_' 
    AND EXISTS {
        MATCH (c1)-[df2:DF_C]->(c2)
        WHERE NOT df2.Type STARTS WITH 'CASE_'
    }
)

RETURN c1, df, c2;

5. In the search bar, run the phrase BPIC17.

This query produces the aggregated EKG showing directly-follows relations from the perspectives of Application, Offer, and Workflow, as well as their interactions.