Detection of excise fraud in trading networks Sébastien Beaudoint – Lore Cloots

Excise Fraudsters Gotta Catch 'Em All #SFBL14 #Excise

SAS Forum 09 October 2014

Outline Introduction Detection of excise fraud in trading networks

Outline Introduction Detection of excise fraud in trading networks

Introduction

 Federal Public Service Finance: 23.409 people Customs & Excise: 3.659 people Investigation & Research: 300 people

Dataminers: 2 people

Outline Introduction Detection of excise fraud in trading networks

Problem definition Excise products Decided at European Level Alcohol (Beer, Wine, Sparkling wine, Spirituous, Intermediate products) Energy products (Petrol, Electricity) Tobacco Minimal rate decided at European Level

Final rate decided by every member state

National excise products Decided by every member state For Belgium Non-alcoholic beverages (except milk) Coffee

Problem definition Rate difference per member state Beer

Wine

Problem definition EU legislation Movements under suspension of excise duty (= duties not yet paid)

Problem definition Evasion at the end of the duty-suspension No payment

Payment in a wrong country (“Excise low level”)

Problem definition Evasion at the end of the duty-suspension Smuggling No document

Multiple use of one document

Problem definition EMCS (Excise Movement Control System) E-AD (Electronic administrative document)

Problem definition: movement overview

Problem definition: project Observation: Smugglers are interconnected Use of social network analysis concepts in the project

Problem definition: project Approach: Risk score Specification of each Belgian company working in the system Number of shipments Number of shipments to low/medium/high excise rate level country Status of shipments Stock

Social Network variables Number of shipments with suspicious company Centrality, betweenness …

E-miner model Decision tree Neural network …

Outline Introduction Detection of excise fraud in trading networks

Data preparation EMCS (Excise Movement & Control System) Information about movements of excise goods under suspension of excise: Excise Product Consignor Consignee Quantity Alcoholic strength Degree Plato (°P, for beer) …

AC4 Information about payment of excise products Excise Product Declarant Quantity in tax base value (depending on excise products: Hl, HlAlc, Hl°P)

Data preparation: Consolidation One line per company ± 2.600 consolidated variables Authorization information Global consolidation Last year of activity consolidation Absolute and relative variables …

Data preparation: problems encountered Data evolution in time Application evolved so the data available also evolved

Missing data & data coherence Solutions Including evolution information in the consolidation Trying to reconstruct missing or incoherent information Including variables indicating data problems (might be a fraud indicator)

Outline Introduction Detection of excise fraud in trading networks

Data exploration In reality, many cycles of data preparation - data exploration Help to check potential use of calculated variables Help to understand some input data

Variable selection Best variables to determine the target variable

Outline Introduction Detection of excise fraud in trading networks

Modeling Model prediction type: decision, ranking, estimates Import input data, define variable types and roles

Divide input data into training and validation samples Define specific models and parameters Compare results

Modeling: models comparison ROC curve

Modeling: problems encountered Quantity & Quality of target data Few positive fraud historical info (± 1 %) No fraud status is poor assumption One model but different fraud behaviours

Solutions Oversampling, under sampling, SMOTE (Synthetic Minority Over-sampling Technique) Probability Matrix and Cost Matrix Creation of different models Going back to fraud history to gather more information …

Outline Introduction Detection of excise fraud in trading networks

Deployment: web application

Deployment: web application

Deployment: web application

Deployment: web application

Deployment: web application

Deployment: web application

Deployment: web application

Deployment: web application

Deployment: web application

Questions?

Thank you for your attention

Questions?

SAS Forum

Twitter Contest – Tweet to win prizes!

3. In which country do you pay the least excise duties for beer?

Belgium

France

Ireland

Prizes to win:

Tweet your answer:

1st prize: a ticket for Analytics 2015 2nd prize: a book of Prof Bart Baesens: “Analytics in a big data world” 3rd to 30th prize: chocolates with pepper

Example: @spicyanalytics 3C

Winners will be contacted post-Forum !

Start of your tweet

Copyright © 2014, SAS Institute Inc. All rights reserved.

Question #

Your answer