Presentation is loading. Please wait.

Presentation is loading. Please wait.

Mobile phone data Belgium State of affairs, datasets, use cases

Similar presentations


Presentation on theme: "Mobile phone data Belgium State of affairs, datasets, use cases"— Presentation transcript:

1 Mobile phone data Belgium State of affairs, datasets, use cases
Marc Debusschere, Statistics Belgium Madrid, WP5 meeting, 7-8 June 2017 Statbel.fgov.be

2 Overview 1. Present state of affairs 2. Datasets 3. Use cases: target statistical outputs 4. Data processing

3 Project Statbel/Proximus/Eurostat
What went before Started December 2015 Several datasets made available for population and tourism 10 publications on analyses and business model Present situation March 2017: meeting on mobile phone data for statistical and commercial use cases, canceled last minute embargo (temporary?) on all data deliveries, including commercial ones, while auditing; reasons only informally communicated What we got out of it Hugely increased technical knowledge and feel for data Concrete idea about use cases

4 For instance … Belgium: population density per km² based on mobile phone data (left) and 2011 Census (right).

5 Other projects & contacts
Academia: UCL/ULB dynamic population mapping using Orange dataset (1 year, CDRs, ) Regional authorities IWEPS (Wallonia): dynamic population mapping (contact 29 June, also intend to use Orange data, maybe paying) Flanders: Dept. Environment Potential users (also producing data!) E.g. Min. Transport, Federal Planning Office, Central Council for the Economy, Belgian Railway company, … DG Regio project on cross-border activity Last but not least: WP5 Mobile phone data!

6 Data situation desperate but not serious …
Proximus Unclear at his moment … Beyond statistics: own exploitation threatened Key: privacy issue, reputation fears Telenet (ex-Base) No reaction to concrete proposal (November 2016, but prior contacts too) Very legalistic, risk-aversive outlook (cfr NDA catch-22) Orange (ex-Mobistar) Seems to provide data to others (ULB - free, IWEPS - paying?) But no reaction to concrete proposal (September 2016)

7 The way forward? Defining use cases Implicating stakeholders
Specify statistical product (including validating or supplementing statistics) Specify data needed Solve technical and legal issues Implicating stakeholders Informing and mobilising potential users Increasing awareness of issues at policy level Legal initiatives General principle (“all your databases are belong to us”) Domain-specific approaches (e.g. tourism, transport, demography, census, …)

8 Datasets received (Proximus) 1
Device counts per Voronoi cell for points in time Purpose: exploration, focus on estimating present population and its dynamics Not using CDRs but signalling data (10 times as frequent) Some 11,000 cells for Belgian territory (averaging 2,8 km², but diverse!) One weekday (Thursday 8 Oct. 2015) & one Sunday (11 Oct. 2015) Recorded every 15’, 96 measurements for each day See description at

9 Datasets received (Proximus) 2
Proximus SIMs observed outside Belgium during 6-month period Purpose: explore tourism trips with destination outside Belgium, confront them with official survey-based statistics Not using CDRs but signalling data (10 times as frequent) 6-month period April 2016 to September 2016 See description at

10 Datasets (Proximus) 3 requested but not (yet) received
Device counts for appr. 11,000 Voronoi cells extrapolated for local market share for each day of the most recent 12-month period, for 45 points in time per day (or, alternatively, for every 15’: 96 points in time) Purpose: improving census results for living place and workplace (static aspect) In principle to be repeated at the beginning of every year Somewhat over 180 million records (or 385 million in the second scenario)

11 Datasets (Proximus) 4 requested but not (yet) received
Cross table of 11,000 most likely living cells x 11,000 most likely work cells based on individually tracked mobile devices in October 2016 Purpose: amend matrix living place-workplace based on administrative data (dynamic aspect) October is most ‘normal’ month, no school or other holidays Record consists of living cell, working cell, number of devices and shapefile of both cells Anonymous, but some cells will have value=1 or low Theoretically 121 million records, but most with value=0 (when nobody living in a particular cell ‘goes to work’ in a particular other cell)

12 Target statistical outputs
Census population by living place & workplace Now based on administrative sources with flaws To be amended using mobile phone data with different flaws By determining most probable living and working place Matrix living place x workplace x X Dynamic population mapping Other Mobility, tourism, time use, circular migration, …

13 Data processing None at this moment … no new data available!
Preparing nevertheless: Trying to solve Proximus issue (Privacy Commission, policy level, …) Contacts with others (ULB, IWEPS) who seemingly have access to data (from Orange) Contacts with data science and analytical capacity in academia and public administration (federal and regional)

14 Comments? Questions?


Download ppt "Mobile phone data Belgium State of affairs, datasets, use cases"

Similar presentations


Ads by Google