DatasetsEvaluation

Aggregation Steps - Datasets

This document outlines a series of aggregation steps used for aggregating EKGs of different datasets and the resulting performances.

---

Soccer (Italy league)

Steps:

step1 = AggrStep(aggr_type="ENTITIES", ent_type="playerId", group_by=["role"], 
            where="birthYear > 1990", attr_aggrs=[])

step2 = AggrStep(aggr_type="ENTITIES", ent_type="pos_orig", 
            group_by=["zone"], where=None, attr_aggrs=[])

step3 = AggrStep(aggr_type="EVENTS", ent_type=None, where='matchId = "2575959"', 
            group_by=[log.event_activity, "teamId", "playerId"], 
            attr_aggrs=[AttrAggr(name=log.event_timestamp, function=AggregationFunction.MINMAX),
                    AttrAggr(name="teamId", function=AggregationFunction.MULTISET)])

step4 = AggrStep(aggr_type="EVENTS", ent_type=None, where='matchId <> "2575959"', 
            group_by=[log.event_activity, "teamId", "playerId"], 
            attr_aggrs=[AttrAggr(name=log.event_timestamp, function=AggregationFunction.MINMAX),
            AttrAggr(name="teamId", function=AggregationFunction.MULTISET)])

Performance:

Step	#Init Events	#Final Events	#Init Entities	#Final Entities	Query Exec Time (s)
Step1	647300	647300	560	238	0.09311890602
Step2	647300	647300	238	231	0.04447603226
Step3	647300	645687	231	231	0.2441709042
Step4	645687	0	231	231	12.88117909
Finalization	0	0	231	0	0.2358920574
Rel Inference	0	0	0	0	9.124970913

---

Order Management

Steps:

    step1 = AggrStep(aggr_type="ENTITIES", ent_type= "customers", group_by=["id"], 
            where=None, attr_aggrs=[AttrAggr(name='id', function=AggregationFunction.MULTISET)])

    step2 = AggrStep(aggr_type="ENTITIES", ent_type= "orders", group_by=[ekg.type_tag], 
            where='price >= 1000', attr_aggrs=[AttrAggr(name='price', function=AggregationFunction.AVG)])
    
    step3 = AggrStep(aggr_type="ENTITIES", ent_type= "employees", group_by=['role'], where=None, attr_aggrs=None)

    step4 = AggrStep(aggr_type="EVENTS", ent_type= None, group_by=[log.event_activity, "customers", "orders"],
            where=f'{log.event_timestamp} >= datetime("2024-03-31T00:00:00")', attr_aggrs=[]) # I want to check all the orders after Easter
    
    step5 = AggrStep(aggr_type="EVENTS", ent_type= None, group_by=[log.event_activity, "customers", "orders"], 
            where=f'{log.event_timestamp} < datetime("2024-03-31T00:00:00")', attr_aggrs=[]) # I want to check all the orders before Easter

Performance:

Step	#Init Events	#Final Events	#Init Entities	#Final Entities	Query Exec Time (s)
Step1	21008	21008	10840	10825	0.08737683296
Step2	21008	21008	10825	9363	0.06592798233
Step3	21008	21008	9363	9345	0.002074956894
Step4	21008	20851	9345	9345	0.1214411259
Step5	20851	0	9345	9345	1.024273872
Finalization	0	0	9345	0	0.1178059578
Rel Inference	0	0	0	0	1.039471865

---

Procure-to-Pay (P2P)

Steps:

step1 = AggrStep(aggr_type="ENTITIES", ent_type= "material", group_by=[ekg.type_tag],where='Delivery_Date_EKPO_BEDAT < dateTime("2023-10-24T09:30:27.930832")', attr_aggrs=[AttrAggr(name='Quantity_EKPO_MENGE', function=AggregationFunction.AVG), AttrAggr(name='Net_Price_EKPO_NETPR', function=AggregationFunction.SUM)])

step2 = AggrStep(aggr_type="ENTITIES", ent_type= "material", group_by=[ekg.type_tag], where='Delivery_Date_EKPO_BEDAT >= dateTime("2023-10-24T09:30:27.930832")', attr_aggrs=[AttrAggr(name='Quantity_EKPO_MENGE', function=AggregationFunction.AVG), AttrAggr(name='Net_Price_EKPO_NETPR', function=AggregationFunction.SUM)])

step3 = AggrStep(aggr_type="ENTITIES", ent_type= "purchase_requisition", group_by=['Purchasing_Group_EBAN_EKGRP'], where=None, attr_aggrs=None)

step4 = AggrStep(aggr_type="EVENTS", ent_type= None, where=None, group_by=[log.event_activity, "material"], attr_aggrs=[])

Performance:

Step	#Init Events	#Final Events	#Init Entities	#Final Entities	Query Exec Time (s)
Step1	14671	14671	8616	6960	0.0979719162
Step2	14671	14671	6960	6320	0.05964684486
Step3	14671	14671	6320	5393	0.04818797112
Step4	14671	0	5393	5393	0.6098949909
Finalization	0	0	5393	0	0.0914850235
Rel Inference	0	0	0	0	0.1724762917

---

Logistics

Steps:

step1 = AggrStep(aggr_type="ENTITIES", ent_type= "Container", group_by=["Status"], where=None, 
        attr_aggrs=[AttrAggr(name='id', function=AggregationFunction.MULTISET), AttrAggr(name='Weight', function=AggregationFunction.AVG), 
        AttrAggr(name='Amount_of_Handling_Units', function=AggregationFunction.MINMAX)])

step2 = AggrStep(aggr_type="EVENTS", ent_type= None, where=None, group_by=[log.event_activity, "Customer_Order"], attr_aggrs=[AttrAggr(name=f'{log.event_timestamp}', function=AggregationFunction.MINMAX)]) 

step3 = AggrStep(aggr_type="ENTITIES", ent_type= "Vehicle", 
        where='Departure_Date > datetime("2023-12-31T23:59:59")', group_by=[ekg.type_tag], attr_aggrs=[]) # I want to check all the orders after Easter

Performance:

Step	#Init Events	#Final Events	#Init Entities	#Final Entities	Query Exec Time (s)
Step1	35413	35413	13888	11889	0.06797218323
Step2	35413	0	11889	11889	1.65262413
Step3	0	0	11889	11823	0.06576704979
Finalization	0	0	11823	0	0.1540901661
Rel Inference	0	0	0	0	0.5269031525

---

Tip

These aggregation steps are just examples. Thanks to the modular nature of the approach, you can combine and adapt them to construct a variety of queries over entity and event data.