HortiCube: A Platform for Transparent, Trusted Data Sharing in the Food Supply Chain Igls-Forum, 18 February 2016 Jack Verhoosel, Michael van Bekkum, Tim Verwaart jack.verhoosel@tno.nl
Motivation: data sharing issues Traditionally limited data sharing: § Sensitivity for competition § Risk of negative publicity § Specific incentives and limited reciprocity § Government does not share data easily, due to issues with ownership and other legal issues 2
Motivation: data availability issues § Available statistics do not present the level of detail required for management decisions ● E.g., product categories: “Pears” versus varieties like Beurré, Williams etc. § Available data are not timely, e.g. export statistics § Growers and traders lack data about consumer trends and how the products are used and appreciated 3
But……there is a need for market insight!! § Market orientation of the horticulture sector can be enhanced: ● Growers go from production-oriented to market-oriented ● Traders start sharing their data, e.g. for better financial management § Towards data based supply planning § From reactive to proactive management 4
And…...relevant market data is there!! § Production: ● Areas/capacities, actual production, forecasts, FADN § (International) trade: ● Import/export statistics, phytosanitary certification requests, stocks, expected arrivals at (air)ports § Consumption: ● Market research, retail sales data, social media § External factors affecting supply and demand: ● Weather conditions, exchange rates, energy prices § News media: ● News about health aspects; food safety incidents 5
Big Data challenges and opportunities § Semantic heterogeneity (in product classifications, etc.) ● semantic integration of data from a diversity of sources § Analytical methods to interpret the data are advanced ● applying them is possible, but not straightforward § SMEs have limited capacity to invest in the IT required for Big Data applications, ● so sharing of investments in this would be welcome § Social media linking as a data source: ● data on consumers’ appreciation and applications of products are abundant in social media ● growers’ communications in social media indicate future supply 6
The BigT&U project § A Dutch public-private partnership § Infrastructure for market information § Access a variety of open data sources § Collect and structure data from social media § Provide semantic mappings between classifications of products § Offer a uniform interface with standard classifications § Boost applications of market data § Enhance market orientation & data based supply planning § Minimize investment for individual SMEs § Privacy and security mechanisms § Enabling of big data analysis over multiple combined data sources 7
8
The HortiCube platform 9
Semantics challenge: mapping ontology § Mapping of similar terms in different data sources § Alignment of food product identification § GPC coding with local/regional classifications § HortiCube contains mapping ontology to achieve this! 10
Mapping product codes of GPC, WAPA, KCB hoofdgroep groep hoofdvarieteit KCB code KCB naam MIT database KCB naam GroentenFruit Huis WAPA NH Variety WAPA ZH Variety GPC Value Code GPC Value Title Hardfruit Appelen 2500 Appelen, alle soorten Appel, alle soorten Total Total 30000881 DESSERT Hardfruit Appelen 2525 Appelen, diversen Appel, overige Other Other 30002515 UNCLASSIFIED Hardfruit Appelen 30002518 UNIDENTIFIED Hardfruit Appelen 30000720 COMBINATION Hardfruit Appelen Alkmene 30015000 ALKMENE Hardfruit Appelen Alkmene Cevaal Hardfruit Appelen Altess Hardfruit Appelen Appelmix Hardfruit Appelen Appels (diversen) Hardfruit Appelen Bellefleur 30015018 BELLE FLEUR DOUBLE Hardfruit Appelen Benoni 2510 Appelen, benoni Appel, Benoni Hardfruit Appelen Bloemee zoet Hardfruit Appelen Braeburn 2517 Appelen, breaburn Appel, Braeburn Braeburn Braeburn 30015026 BRAEBURN Hardfruit Appelen Campagne zoet Hardfruit Appelen Canada Grise Reinette Grise du Canada 30015216 REINETTE GRISE DU CANADA Hardfruit Appelen Civni 2571 Appelen, rubens Appel, Rubens 30015036 CIVNI Hardfruit Appelen Collina Hardfruit Appelen Contento Hardfruit Appelen Cox 2520 Appelen, cox's orange Appel, Cox's Orange Pippin Cox Orange 30015039 COX’S ORANGE PIPPIN Hardfruit Appelen Cripps Pink Cripps Pink Cripps pink 30015041 CRIPPS PINK Hardfruit Appelen Crown Apple Hardfruit Appelen Dalinbel Hardfruit Appelen Dalinco Hardfruit Appelen Delbarjubilee Hardfruit Appelen Delblush 30015052 DELBLUSH Hardfruit Appelen Delcorf 30015053 DELCORF Hardfruit Appelen Dijkmans 30015062 DYKMANNS ZOET Hardfruit Appelen Discovery 2522 Appelen, discovery Appel, Discovery 30015059 DISCOVERY Hardfruit Appelen Diversen (hardfruit) Hardfruit Appelen Early Scarlet 30015817 EARLY Hardfruit Appelen Elan 30015064 ELAN Hardfruit Appelen Elise 30015065 ELISE Hardfruit Appelen Elsmi Hardfruit Appelen Elstar 2530 Appelen, elstar Appel, Elstar Elstar 30015068 ELSTAR Hardfruit Appelen Elstar * 2530 Appelen, elstar Appel, Elstar Elstar 30015197 RED ELSTAR 11
Security and privacy: token-mechanism § Security mechanism now incorporated is fairly basic ● Token-based dual password-oriented § Currently in development: application of mechanisms for ● role-based accessibility per specific data source ● user-specific accessibility for sets of data elements. § Privacy is ensured by: ● Anonymisation of data ● Aggregation of data 12
Apple/Pear use case § Global apple/pear 2006-2014 production data and production forecast for 2015 (WAPA) § Dutch apple/pear export data to EU countries (KCB) § WAPA ontology and KCB ontology § Mapping ontology of apple/pear varieties § Company-specific data is incorporated § Token-based user/password mechanism 13
HortiCube and Apple/Pear use case Google Visualisa+on Applica.on example Javascript SPARQL interface Hor+cube Common ontology Apache Jena Fuseki triplestore Mapping Apache Marmo?a WAPA Triples KCB Triples Mapping-ontology triplestore KCB-ontology LODRefine WAPA-ontology LODRefine .CSV .CSV WAPA Production 2006-2014 + Forecast 2015 KCB Export 2010-2014 14
Big data demo application HortiCube interface enables development of demo application that provides answers to the following important information questions from apple/pear growers: § What is the trend in apple/pear production per country and what are the consequences for the export distribution to other countries? § What is the relation between monthly apple/pear export and local weather conditions in this time period? 15
Big data analysis question and result 16
Conclusions § A shared business incentive is very important for stakeholders to share data! § This can be found in specific sources of data with direct added value to the stakeholders, e.g. timely data! § Mapping of product classifications is a laborious activity ● However, once done it can be reused many times! § Opening up data sources and making them linkable via the HortiCube is feasible in a relatively small amount of time ● Updating these data sources therefore becomes possible, especially when the incorporation step can be mostly automated. 17
Thank you for your attention! 18
Recommend
More recommend
