View
217
Download
0
Category
Preview:
Citation preview
Data analytics in the Spanish Tax
Agency
Analytics Experience 2016
November 7-9, Rome
José Borja Tomé
Subdirector General de Tecnologías de Análisis de la Información e Investigación del Fraude
Analytic tools usage
# Users
Analytic capabilities
Aggregated analysis
Simple network Analysis
Advanced models
Teseo
Market tools
Zújar (DW) Search engines
Standard reports Individual query
Zújar. The data warehouse
# Entities
# Visualizations
# Users
Governance: a role of user managers+
IT staff
Governance
Of
Information
Completeness
Semantic
Quality
Security
Integration
MIDAS - ROI
Applications to be included in ROI
2009-2015
•Filters (rules)
•Results: •Inclusion
•Denial
•Other information
Starting point
77% to check
Training
Execution
•Filters, other information
New applications to be included in ROI
•Results •Inclusion + motives
•Denial + motives
•To check + % of risk
With analytic system
45% to check
Additional benefits:
•Analysis and enrichment of filters
•Homogenization of decisions
•Ordering of files by risk
•Explicability of the model
•Continuous control, not subject to
campaigns
• Initially:
– Identify clear and well defined cases
– Limited scope projects
– Small number of expert business users
– Reduced governance: – Basic security
– Basic interfaces with business applications
• Evaluate the results and if they are satisfactory…
• More complex cases, study corporate adoption, increase governance and
scalability
IT Analytic project leaders
Model factory
Business Analytic project leaders
IT operational Apps staff
Organizing advanced analytic projects
ELECTRA
• Some groups of taxpayers
have an anomaly in the
distribution of their internal
versus external commercial
relations.
• In some cases these groups
are created with a
fraudulent purpose,
especially when they have
close bonds.
ELECTRA. Expansion of
relationships
Feedback Decision taken by the
official
Finding user cases of predictive
analytics. MIDAS – IVAN
(VAT refund claims for non established)
HISTORIC DATA •IVAN Files •IVAN declared invoices •DUAs( SADs) ofExport/Import •Claims against taxt decisions •Professional fuel refunds •Mod. 340- Invoices •… OBJETIVE •Decision taken by the official
MIDAS Decision
Rules
Weights
Variables
Predictive model
NEW FILES •File data •Declared invoices •DUAs (SADs) of Export/Import •Claims against taxt decisions •Professional fuel refunds •Mod. 340- Invoices •…
IVAN management
system
Operation Feedback Training
The case
The data
The advantage
Web scraping
• Although advanced analytics and big data are important, most needs can be satisfied with less sophisticated technologies (decisions motivated in true facts)
• Access for everybody (citizen data scientists) needs governance
• IT/Business Analytic project leaders + Model factory
• A bimodal approach simplifies the adoption of technologies
• Operational systems must integrate the results of the models
• Internet provided useful information in targeted areas although identification and uncertainty must be dealt with
Conclusions
Data analytics in the Spanish Tax
Agency
Analytics Experience 2016
November 7-9, Rome
José Borja Tomé
Subdirector General de Tecnologías de Análisis de la Información e Investigación del Fraude
Recommended