call download fax letter pdf search x chevron
Stage Image
Nowcasting Inflation:
Machine Learning for Automated Product Classification

The FS Hackathon in cooperation with the European Central Bank is a 25-hour Hackathon taking place at Frankfurt School of Finance and Management.

Up to 40 participants will take on this hacking challenge. As part of a team of passionate data scientists, you will develop and present innovative solutions for an automated classification of thousands of individual products into pre-defined categories. Products are scraped from the web pages of major German retailers

Are you interested in taking on this challenge? Register here for a chance to take part in the FS Data Science Hackathon with the ECB. The event is free of charge, but please keep in mind that participation is subject to a pre-selection process based on skills and availability.

Date and Time

20 September 2019, 18:00 to
21 September 2019, 20:00

Registration deadline

5 September 2019

Challenge Description

To use the full potential of web-scraped product price data for monitoring and nowcasting inflation rates, millions of products available on the internet must be classified using a common classification system. Given the number of different products found online, this is a task impossible for humans. So we need a solution that automatically classifies products found on different websites using a common classification scheme.

The challenge of this hackathon is to classify already web-scraped supermarket products into a pre-defined classification system (the “COICOP” classification as used by statistical offices in the euro area). On the basis of a dataset classified by experts, you will be challenged to build a classifier that achieves high accuracy in the classification of previously unseen products.

Background

Price stability is at the core of the mandate of the European Central Bank (ECB). Accordingly, the ECB has a long tradition of monitoring and forecasting inflation and studying its determinants. Fundamentally, however, inflation is determined by the behaviour of millions of price setters, so the ECB also has a natural interest in studying the prices of individual goods. Since the explosion of the internet it has become possible to scrape information about the price of goods sold online and about the prices of services such as restaurants, hotels and flights. Never before have we had at our fingertips such a vast amount of information on the behaviour of individual prices. This new and, until now, unavailable data source has attracted the interest of several areas of the ECB, which is currently investigating the usefulness of web-scraped data for various purposes. Web-scraped data on individual product prices can provide invaluable input for monitoring and nowcasting inflation, both of which are core activities of the ECB.

Nowcasting

The following figure illustrates the process of nowcasting. It consists of 4 basic steps:

1.    Acquiring data: web crawling from various pages
2.    Raw data parsing: converting product information to structured tables
3.    Processing structured data:
       a.    data quality
       b.    intelligent classification (scope of this hackathon)
4.    Creating time series to nowcast inflation

Prizes

The two winning teams of the hackathon will be invited to present the results at the ECB immediately following the conference “Inflation in a changing economic environment"on 24 September 2019.

European Central Bank in Frankfurt

European Central Bank in Frankfurt

Judging Criteria

The solutions will be evaluated on the basis of four different criteria:

Accuracy

Number of products correctly classified in an evaluation data set.

Generalisability

The hackathon will use a limited number of websites and one language (German), but the plan is to use web scraping to cover more websites, languages and product groups. A solution that could be extended to other websites, without needing a large training data set or relying on the specific structure of the website at hand, would be highly welcome. To meet these criteria it is important that you are able to demonstrate the potential and limitations of the proposed solution.

Innovation

Novelty of the approach: this may involve, but is not limited to, using transfer learning approaches, automl (e.g. using Gridsearch), online machine learning approaches to avoid scheduled/batch retrainings, other data sources, etc. It is important that you demonstrate the potential and limitations of such an approach, in particular with regard to accuracy.

AI Monitoring and Business Process Integration

How can we ensure that products are correctly classified in production? How can we check what the classifier is doing? How can we have a system that alerts us when the classifier is significantly off track and/or requires updating? How and in what form could experts give feedback to the classifier (e.g. completely wrong, nearly right (if some COICOP levels are right, but not all), correct)? How can the classifier signal to an expert that it is unsure about a suggested classification of a product and that this should be checked by an expert?

Participants

Registration is open to anyone with the necessary skills to complete the challenge. FS students are encouraged to participate.

The maximum number of participants is 40, so there will be a pre-selection phase to determine who takes part in the hackathon. The organisers will particularly consider diverse and multi-disciplinary teams with a wide range of experience and skills.

Remember to bring your electronics: laptop, phone and chargers.

If you use commercial software, you will be required to hold the relevant licenses. We explicitly do not provide any software.

Fee

Registration is free but you will be subject to a pre-selection process.

Teams

You will be assigned to a team, with a maximum of five participants per team.

The organisers will strive to create diverse teams with a wide range of experience and skills across the different team members.

Participation Agreement

Participation

The Competition is free and without any purchase obligation. The Competition is open to everyone who have the necessary skills.

Employees of FS or ECB, their affiliates, and members of their families, as well as anyone who has directly or indirectly contributed to the conception, organization or realization of the Competition are ineligible to take part in the Competition.

Each registration in the Competition is individual, and only one registration per person is allowed during the entire duration of the Competition.

Registration Process

Every person wishing to participate can register online on the dedicated website.

Each candidate guarantees that the information entered during his/her application to the Competition is complete and accurate. Any false, incorrect or incomplete information will automatically result in the application being rejected.

The organiser reserves the right to reject an application at any time if the profile of the participant does not comply with the Competition standards.

Teams Formation

FS and ECB will form teams of two (2) to five (5) participants. One participant cannot be a part of several Teams.

Compensation

Participant waives all its claims regarding compensation against the organiser due to its participation to the Competition (in particular with regards to cost incurred for participation such as the transport, accommodation, etc.) with the exception of drinks and meals which will be supplied by the organiser during all the duration of the Competition.

Modification of the Competition

The organiser reserves the right to shorten, extend, modify or terminate the Competition, or a part of the Competition as a result of a force majeure event without incurring liability as a result thereof. In such case, the organiser will provide notice thereof to the Participants by any means (including email and/or publication on the website (by its choice), and, if the need arises, will communicate to them the new rules applicable or the new ending date of the Competition, as the case may be. Each Participant may notify by e-mail to the organiser that s/he refuse the new conditions of the rules, in which case s/he will be excluded from participating, which s/he expressly accepts. If no refusal is notified to the organiser within 72 hours, the new conditions of the rules will be deemed accepted by Participants.

In case of a modification to the rules, cancellation, interruption of the Competition, a reduction or an extension of its length, the organiser waives all liability and participants will not be entitled to any compensation.

Fraud

The FS may cancel all or part of participations in the Competition if it appears that fraud has occurred in any form whatsoever, in particular computer fraud, during the Competition. The FSr reserves, in this case, the right not to award prizes to fraudsters, to disqualify the concerned solution project and/or to bring actions against the authors of these frauds. The FS shall not incur any responsibility towards the Participants because of the committed frauds.

Furthermore, the organiser cannot be liable in case of damage, whether material or immaterial, caused during the Competition and thereafter, to the Participants, to their computing equipment to the data and to the consequences which can ensue from it on their personal, professional or commercial activity.

Exploitation Rights and Intellectual Property

  • Any participant irrevocably grant to the ECB the exclusive right to use for all documents, data, information, scripts, tools, presentations and other work results that he/she produces in performing the hackathon (‘Work results’), in each case as from the time when such rights are protected by law. The ECB’s right of use shall not be subject to any restriction in terms of time, place or application and it shall cover all forms of use known at the time of entering into this Contract.
  • The right to use granted in above shall include, without being limited to, the rights to copy, distribute, process, modify, further develop and/or dispose of the Work results or create derivative works based on the work results. The granting of rights also refers to types of use unknown at the present time, whereby the participants mandatory rights provided for under the applicable law (e.g. with regard to withdrawal or to adequate remuneration), if any, shall remain unaffected.
  • Only the ECB shall be entitled (but not obliged) to apply to register worldwide intellectual property rights for the Work results in its own name and shall become the owner of all such intellectual property right applications and all resulting intellectual property rights.
  • The ECB may transfer the rights of use or may grant rights thereon (sub-licences) to third parties in whole or in part.
  • Mandatory rights (in particular moral rights), if any, shall remain unaffected
Image Rights

The Participants may be filmed or photographed during the event. By joining the Competition, the Participants agree to the use and the distribution of their image by ECB and FS, including for promotional event occurring after the Competition. The production and distribution of films and photographs of the event will not lead to any compensation of the Participants.

Protection of Personal Data

The collection and processing of personal data relating to the participant by the organiser is primarily intended to ensure that the Competition will be properly completed and in particular to allow the organiser to contact the winner and to award them the prizes in an efficient way and, if the cases arises, to publish the list of the winners.

Agenda

September 20
18:00 Doors open and registration
18:30 Welcome note: Prof. Dr. Nils Stieglitz and Vahe Andonians, FS
18:40 Event introduction: Dr. Jörn Franke and Dr. Lukas Henkel, ECB
19:00 Warm-up session: Ines Dolić, FS
19:15 Technical introduction: Dr. Michael Menzel, Google
19:45 Start hacking
20:30 Dinner buffet
22:00 Snacks all through the night
22:00 Break out session: Rojan Gharadaghy, Google, room 3.01
22:30 Continue coding until the next day
September 21   
07:30 Breakfast buffet
08:00 Continue coding and working on projects
10:00 Coffee and snacks
11:30 Open lunchtime, Fine Dining (ground floor)
14:00 Continue coding and working on projects
16:30 Hand in projects
17:00 Pitches five minutes per team + short Q&A
17:50 Networking while projects are being judged
18:30 Awards
19:00 Goodby drink
20:00 Close

Mentors

A team of experts will be available from the start to help you develop your solution as far as possible. Mentors do not participate in the evaluation and award process. Their role is completely neutral.

Vahe Andonians

Frankfurt School of Finance & Management
Data Science Expert
Entrepreneur
Senior Lecturer

Dr. Jörn Franke

European Central Bank
DG Information Systems
DIV Data Analytics Domain Services
IT Expert

Dr. Lukas Henkel

European Central Bank
DG Economics
DIV Prices & Costs
Economist

Prof. Dr. Olaf Stotz

Frankfurt School of Finance & Management
Professor of Asset Management and Pension Economics

Georgios Prassas

European Central Bank
DG Information Systems
DIV Data Analytics Domain Services
Senior Project Manager

Alexandros Melemenidis

European Central Bank
DG Economics
DIV Business Cycle Analysis
Research Analyst

Daphné Aurouet

European Central Bank
DG Information Systems
DIV Data Analytics Domain Services
IT Specialist - Data Scientist

Camila Carlos Ballerini

European Central Bank
DG Economics
DGE Data Integration Team
Analyst – Data Scientist

Paluszek Andrzej

Directorate General Secretariat to the Supervision Board
Supervision Quality Assurance Division
Supervision Analyst – Data scientist

Dr. Alexander Del Toro Barba

Google
Machine Learning Specialist

Dr. Michael Menzel

Google
Cloud Customer Engineer - Big Data & ML
 

Judges

Dr. Chiara Osbat

European Central Bank
DG Economics
DIV Prices & Costs
Adviser

Dr. Christoph Schaper

European Central Bank
DG Information Systems
Head of Division Innovation & Architecture Services

Dr. Martin Eiglsperger

European Central Bank
DG-Statistics
DIV Monetary & Economic Statistics
SCT Economic Statistics
Senior Economist-Statistician

Rojan Gharadaghy

Google
Account Executive - Google Cloud

Location

Frankfurt School

AI-Lab

Tower 4A

Adickesallee 32-34

60322 Frankfurt am Main

Deutsche Bahn Event Ticket

It couldn’t be any greener. Travel by train to your event for as little as € 99 with 100 % green energy.
Learn more.