Coopetitive Data Warehouse: a case study Andrea Maurino, Claudio Venturini, Gianluigi Viscusi DISCo - Dip. di Informatica, Sistematica e Comunicazione Università di Milano Bicocca viale Sarca 336/14, 20124, Milano (Italy)
Jun 25, 2015
Coopetitive Data Warehouse: a case study
Andrea Maurino, Claudio Venturini, Gianluigi ViscusiDISCo - Dip. di Informatica, Sistematica e Comunicazione
Università di Milano Bicoccaviale Sarca 336/14,
20124, Milano (Italy)
Index
• The problem• Co-opetition, who are you?• Theoretical background• A methodology for designing coopetitive
information systems• The AOPUnoLombardia case study• Evaluation• Conclusion and future work
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 2
The problem
• In each market firms are always in competition• But, sometime, they need to be aggregated
• Law imposes or encourages such aggregation (lobbying, founding)
• better understand the market • Thus firms sometime need to cooperate
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 3
Co-opetition, who are you?
• Coopetition is a kind of relationship in which firms show competition and cooperation behaviours
• Very well studied in economy• NOT STUDIED IN ICT!
• We studied the problem of designing a coopetitive datawarehouse where firms share their sensitive data
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 4
Theoretical background
• Coopetition can be modelled whit game theory• Let
• A, B two firms• {(S) share – (NS) no share} possibile decision• Ka and Kb amount of shared knowledge• Lv is the loss value due to share information with others• Av is the added value to have information from others
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 5
Theoretical background
• Share, share is a dominant solution for a company x if• Kx-Lv+Av>Kx
• How to measure Av and Lv? (any idea?)• But in Lv is relied on
• Access• Trust• Privacy• Quality
• We reduce the value of Lv
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 6
Theoretical background
• Access control policies;• Extraction of data from organizations• Integration and load into the DW• Presentation of results
• Trust • If you want to freely join in an organization you
trust in it!• Privacy preserving strategies • Quality of shared data.
• Reduced quality reduced value of information
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 7
The architecture
• GVV is a global virtual view integrating existing schema• Data are still in whithin firm borders and they are queried
when needed• It is a virtual operational data store
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 8
• When ETL asks for data they arequeried by the GVV to firms, intregrated on the fly and stored in DW• Anonymous
The methodology
• 2 phases requirement• Concept definition
• Aggrement among firms on WHAT Dw will mange data
• Definition of the global model
• Quality of exposed data • Data aggregation
• Concepts classification• Aggrement among firms on
HOW Dw will manage data• Mandatory Access
control• 5 security levels• Unclassified,
classifed, confident, secret and top secret
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 9
The methodology
• 1. unclassified: concept X can be shared freely.• 2. classified: plain concept X can be shared with a
intermediated quality level• 3. confidential: concept X can be shared in an
aggregate way with a intermediate quality level.• 4. secret: concept X can be shared, in a way to
preserve privacy, and with a• low level quality.• 5. top secret: concept X must not be accessed.
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 10
The methodology
• How reduce quality of data?• Reduce
• Accuracy• Completeness• Consistency• Time related
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 11
Also for antitrust law…
The AOPUnoLombardia case
• AOP UnoLombardia is an organization of fruit and vegetable producers. • 13 grower organizations (GO) including
• Bonduelle and Dimmidisi. • It represents about the 30% of the whole fruit and
vegetables market in Italy.
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 12
The AOPUnoLombardia case
• AOP UnoLombardia wanted to develop a coopetitivedata warehouse for the following goals:• to obtain a unified analysis of selling of goods in term of
price and amount of sold pieces• to obtain a unified view of raw materials bought
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 13
The AopUnoLombardia case
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 14
Evaluation
• May 2011 E.coli bacteria outbreak
• No deads in Italy at all!• Anyway …PANIC!
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 15
Evaluation
• AOPUnoLombardia use our system to evaluate • the reduction of the amount of product type sold to large-
scale retail wrt the the previous year• the reduction of purchase of raw material wrt the the
previous year
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 16
- Average reduction of sold products is about 29%-In some case the number of sold product was reduced by a 38% -only in September 2011 the amount of sold items had the same value ofthe previous year.-while the amount of brought salad is more or less the same!
due to preexisting aggreements
Conclusion and future works
• Coopetition is a kind or relationships in which firms show a cooperative and coopetitive behaviours
• The problem of building a coopetitive informatiom system is a new problem!
• We started by design and impement a coopetitive datawarehouse• Applied to AOPUnoLombardia case study
• Future Work• From a theoretical view point a better definition of
information value • We are currenting extend our system to include other GOs• We are studing how to apply our system in the ortofruit
district in Lombardia including seeding and tools producing firms
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 17
Thanks!
疑问
••• ITIS Lab ••• http://www.itis.disco.unimib.it ••• 18