Data Science in BIS Current activity and prospects Siobhan Carey Chief Statistician March 25 2015
2 • What we are doing • Some reflections
Where we started from Topic specialism Data Science IT / analytics Stats/ inference
4 Key dimensions • Infrastructure • Skills – Data analytics – Data visualisation – Interpretation • Data – Big or otherwise
Topic specialism BIS DATA BIG DATA Data Science IT / analytics Tools Stats/ inference Platform
6 Tools Data visualisation /Mapping D 3, Leaflet, Mindjet, Map. Info, Arc. GIS, QGIS, GIMP, Inkscape, Scribus… Scripting/Coding Active. Python, Active. Perl, Openrefine, Javascript, Gedit, Netbeans, Tortoise SVN, Scala Statistical Analysis SAS, SPSS, Stata, R, Matlab, Vensim, X 13 Arima, SQL … Stuff Office, Browsers Deployed on a scalable virtual machine
7 Skill development • • Small central team Training /exposure New toolbox available to all (90 as of this week) Projects – Trade visualisation – Companies House
8
9 Skills - interpretation
10 What are we hoping for? – New insights from our existing data holding – New ways of presenting data • Easier to absorb, easier to explore – New data that can be harvested relevant to BIS policy decisions • Cheaper, more timely? • Replace existing data? • Supplement existing data?
11 Data Existing BIS data: • Large datasets in HE and FE • Data on businesses • Data linking New Data sources • Big data – some potential projects • Personal data – legal and disclosure issues
12 REFLECTIONS
Current paradigm Farmer • Cultivation • Theoretical base • Known properties - Coverage - Non-response - Known and tested variables • Skills well developed if in short supply • Infrastructure • Process • Security of supply of inputs
New paradigm?
Current paradigm – New paradigm Farmer • Cultivation • Theoretical base • Known properties - Coverage - Non-response - Known and tested variables • Skills well developed if in short supply • Infrastructure • Process • Security of supply of inputs Forager • Lacks statistical properties • Unknown coverage/bias • Doesn’t necessarily have the measures you want/need • Lack of standardisation • Needs alternative theoretical framework • Growing on someone else’s land? • Future supply uncertain
18 Thank you! siobhan. carey@bis. gsi. gov. uk http: //analysis. bis. gov. uk/trademap/ Images 123 RF 5050330 Pavel Mitrofanov, 5050330 loganban 6666493 Silvia Crisman
Скачать презентацию Data Science in BIS Current activity and prospects
5ba0ef43c5979c41384615159619ecad.ppt