20141030 LinDa Workshop echallenges2014 - Linked Data Analytics

Workshop 7d, 30 October 2014 eChallenges e-2014 Copyright 2014 , Ubitech Ltd.
Business Value Creation from Linked
Data Analytics: The LinDA Approach
Anastasios Zafeiropoulos, Eleni Fotopoulou
Ubitech Ltd./R&D Department
Athens, Greece
azafeiropoulos@ubitech.eu

Introduction
• Existence of a wide set of available data sources nowadays with
parallel lack of means to exploit them in an optimal way and realise
advanced analysis.
• Need for proper interconnection of concepts for managing to examine
the relationship among entities represented in different data sets.
• An approach for the extraction of Linked Data analytics is presented,
based on the exploitation of a set of tools for proper transformation and
interlinking of public and private datasets and the realization of analysis
over them.
• The proposed approach targets at enhancing the ability of public and
private sector organizations to provide usable Linked Data, while
offering SMEs the opportunity to perform advanced algorithmic
analysis.

Challenges (1)
• Need for management of structured and un-structured data in multiple
formats that in some cases lack representation based on defined
schemas.
• Data aggregation from distributed sources
• increasing wealth of dataset cross-linkage;
• SPARQL queries cannot readily be executed as their constituent triple
patterns span across multiple datasets.
• Compilation of proper and meaningful datasets to be provided to the
analytics tools
– review the datasets and prepare them in proper format;
– extract knowledge from the data through interlinking, inferences as well as
analytics extraction;
– maintain and update the data regularly;
– elimination of co-references among the available data.

Challenges (2)
• Handle data with different quality characteristics regarding
their accuracy, consistency, timeliness, completeness,
relevance, interpretability and trustworthiness
– set of information quality assessment metrics;
– some of the indicators cannot be automatically assessed;
– data quality assessment is performed only on a small sample of the
data which results in a decrease of the precision of the quality
scores;
• Need to process high volume data in some cases as well
as have the capacity to apply and evaluate the results of
proper algorithms.
• Learning curve for the adoption of Linked Data technologies
from SMEs and public administrations.

Linked Data Analytics: the LinDA approach
• Linked Data: a set of best practices for representing and
connecting structured information on the web.
• LinDA addresses one of the most significant challenges of
the usage and publication of Linked Data, the renovation
and conversion of existing data formats into structures that
support the semantic enrichment and interlinking of data.
• The proposed approach is building upon the collection of
data from available data sources, their transformation in
proper format (e.g. RDF format) and their interlinking for the
creation of extended linked datasets, fed as input in the
analytics extraction process.

The LinDA Approach

Linked Data Analytics – Algorithms
Categorization
• A library of basic and robust data analytic functionality is
provided.
• Design and deployment of workflows for algorithms
execution based on their categorization:
– Classifiers for identifying to which of a set of categories a new
observation belongs based on a training set;
– Clusterers for grouping a set of objects in such a way that objects in
the same group are more similar to each other;
– Statistical and Forecasting Analysis for discovering interesting
relations between variables and providing information regarding
future trends;
– Attribute Selection (evaluators and search methods) algorithms for
selecting a subset of relevant features for use in model construction
based on evaluation metrics;

Integrated Components
• Based on existing open-source platforms for extracting data
analytics.
• Weka open-source tool.
• R project for statistical computing.
• Customized end-user applications for selected business
domains, targeted at reducing the overall complexity in the
configuration of the algorithms and the preparation and
management of the linked datasets.

Analytics Ecosystem Components

Interconnection with LinDA Components

Production of Linked Data Analytics
• Supporting RDF input and output.
• RDF to CSV transformation in the Publication and
Consumption Framework.
• Enriched CSV input loaded to the analytics tool: metadata
for initiating RDF URI, submitted query, analytics process
id, analytics process description, storage options at LinDA
repository.
• RDF output available in the LinDA repository.
• Analytics output available for creation of visualisations
(where appropriate).

Elaborated Ontologies
• FOAF Ontology: describe person activities;
• PROV Ontology: represent and interchange provenance information
generated in different systems and under different contexts;
• SIO Ontology: simple upper level comprised of essential types and
relations for the rich description of arbitrary (real, hypothesized,
virtual, fictional) objects, processes and their attributes.

Pilots & Business Value
• Business Intelligence Pilot
• Media Analytics Pilot
• Environment Analytics Pilot
• Redesign of current business processes
• Reduction in overall complexity and
administration overhead
• Deployment of novel services
• Short Demo

info @
LinDA-project.eu
@LinDA_FP7
+
LinDA-project.eu
LinDAFP7
Thank you! Questions? Anastasios Zafeiropoulos| azafeiropoulos@ubitech.eu
Senior R&D Architect
Ubitech Ltd.| www.ubitech.eu

20141030 LinDa Workshop echallenges2014 - Linked Data Analytics

More Related Content

What's hot (20)

Viewers also liked (6)

Similar to 20141030 LinDa Workshop echallenges2014 - Linked Data Analytics (20)

More from LinDa_FP7 (7)

Recently uploaded (20)

20141030 LinDa Workshop echallenges2014 - Linked Data Analytics