Microsimulation, Big Data and Predictive Analytics

Transcription

1 Microsimulation, Big Data and Predictive Analytics Mark Birkin Professor of Spatial Analysis and Policy, School of Geography, University of Leeds Director, ESRC Consumer Data Research Centre Director, Leeds Institute for Data Analytics

2 ESRC Big Data Network National statistical authorities UK Government departments Research Institutes World-wide data archives Administrative Data Research Network Identifying suitable data Negotiating access Identifying a safe data access setting Social Media Data & Third Sector Data Further announcements soon Phase 1 Phase 2 Phase 3

3 CDRC Data Partners Multilateral data sharing Case studies & academic publications Metadata & provenance Business engagement & awareness raising Providers Partners Prospects Participants

4 CDRC Data Partners

5 Health Simula=on 2011 Census Gender x Age x Ethnicity Gender x Age x Illness Gender x Age x NSSEC Gender x Age x Car ownership Spa$al Microsimula$on (2011) ELSA Wave 5 ELSA Waves 1 to 6 ETHPOP 2011 to 2031 Projec$ons Adjusted Projec$ons Gender x Age x Ethnicity Hybrid Microsimula$on (2011 to 2031) Hazard Model Take account of future ethnic composi=on of the local authority popula=on CHD; stroke; diabetes; cancer; respiratory illness; arthri$s and depression Clark S., Birkin M., Heppenstall A. (2014) Sub regional es$mates of morbidi$es in the English elderly popula$on, Health & Place.

6 Health (2)

7 Health (3)

8 Health (4): App Data Weekly ac$vity readings from the Bounts app: Clocks change Clocks change

9 Transport: Cycling Blue = commuter cycling potential Green = travel to school cycling potential Key: Source: hzp://rpubs.com/robinlovelace/245696

10 Transport: Trains

11 Transport: Trains

12 Transport: Journey Planning Journey to work

13 Infrastructure

14 Infrastructure

15 Energy Zuo et al (2014) Geospa$al Informa$on Science, 17, 3,

16 Retailing/ Consump=on

17 Retailing/ Consump=on

18 Recap on Examples

19 The Fourth Paradigm a new method of pushing forward the fron$ers of knowledge, enabled by new technologies for gathering, manipula$ng, analyzing and displaying data Thousand years ago, science was empirical, describing natural phenomena Last few hundred years, theoretical branch, using models, generalizations Last few decades, a computational branch, simulating complex phenomena Today, data exploration (escience) synthesizing theory, experiment and computation with advanced data management and statistics à new algorithms! (Alex Szalay)

20 Predic=ve Analy=cs Data is growing, but what about your ability to make decisions based on those huge volumes of data? Success will depend on how quickly you can discover insights from all that data and use those insights to drive bezer ac$ons across your en$re organiza$on That s where predic$ve analy$cs, data mining, machine learning and decision management come into play. Predic$ve analy$cs helps assess what will happen in the future. Data mining looks for hidden pazerns in data that can be used to predict future behaviour. Businesses, scien$sts and governments have used this approach for years to transform data into proac$ve insights. Decision management turns those insights into ac$ons that are used in your opera$onal processes. (sas.com)

21 A Note on Synthesis Two defini$ons of Synthe.c: devised, arranged, or fabricated for special situa$ons to imitate or replace usual reali$es made by combining different substances : not natural The second is relevant as well as the first!

22 Three problems for MSM Calibra$on and valida$on? especially in real-$me Behaviour From demographic structure to ac$vi$es and impacts Predic$ve analy$cs robustness and relevance Microsimula=on needs Big Data!

23 Three problems for Big Data Representa$on bias Inferences for the popula$on, not the sample Synthesis Context is always crucial Privacy, confiden$ality, ownership and trust Conflict between granularity and disclosure Big Data needs Microsimula=on!!

24 Conclusions Strong synergy between MSM and Big Data Dynamics over mul$ple scales increasingly important Consider synthe$c popula$ons as synthesising as well as synthesised