If Norfolk is to benefit from the opportunities offered by big data and data science, business and public sector organisations need to work with universities and researchers. To promote such links, the second event of the Meetup Group “Data Science Norfolk” took place on Thursday 11 May 2017 at the historic AVIVA Marble Hall at Norwich’s Surrey Street. The meetup, which is organized by Norwich Business School academics Nikolaos Korfiatis (Senior Lecturer in Business Analytics) and James Cornford (Senior Lecturer in Organizational Behaviour), was kindly hosted by Steven Bacon (Principal Data Scientist) from AVIVA’s Data Science team, The LAB. AVIVA is actively sponsoring the meetup events of Data Science Norfolk, which aim to enhance interaction and sharing of ideas about the exciting new field of data science and analytics across all industries in Norfolk and East Anglia. Participants came from a range of organizations and companies in the region (e.g., Norfolk County Council, Elevate Credit, UEA School of Computing Sciences, Norwich Business School as well as AVIVA and Novartis) had the opportunity to discuss ideas and network over drinks and pizza (as it is customary in every meetup event).
The presentations were kicked off by Steven Bacon who gave an overview of the work done by the data science team in AVIVA. The presentation looked back at the history of data science in the company before describing some of the exciting new things coming in the field of insurance pricing, building on advances in technology such as apps that monitor your driving style. It was interesting to see the historical contrast from the installation of AVIVA’s (then Norwich Union) first Mainframe to the modern data processing equipment used by underwriters and how Norwich’s underwriting unit has evolved to the Data Science team, an interesting transition to observe as Rod Moyse the head of analytics for AVIVA remarked later to the participants. Steve showed some interesting examples demonstrating that as a customer focused organisation, Aviva takes data protection and transparency very seriously.
The next presentation was delivered from NBS’s James Cornford on the business/organizational aspects of running a data driven organization. James gave a political scientists overview of the challenges that data- or evidence-based decision making can create. For example, he noted the dangers of Goodhart’s law – when a measure becomes a target it ceases to be a good measure. James, who is also part of the ESRC Local Government Data Research Centre, hosted at UEA, presented a number of interesting cases of the risk, challenges and dilemmas of data driven decision-making. A powerful example came from adoption of children from care where insight from patterns in large volumes of data need to be combined with very detailed information about individual cases, with the two sources of evidence sometimes pointing in different directions.
The third presentation in the event was delivered by Aristidis (Aris) K. Nikoloulopoulos with the intriguing title: “The copula that ‘killed’ Wall Street and beyond”. Aris has pointed out the limitations of normal copula for modelling high-dimensional financial data and has recommended vine or factor copulas when data display dependence among extreme values and inferences based on multivariate tail probabilities are needed. Aris is a Senior Lecturer in Statistics at the School of Computing Sciences and is notable for his contributions in the field of (vine/factor) copulas. He is one of only a handful of researchers worldwide that use copulas for modelling multivariate discrete response data (e.g., the number of insurance claims for an individual policyholder within both a longitudinal and multivariate context). He is also an avid R programmer having released two packages in CRAN related with dependence modelling techniques.
The final presentation was delivered by NBS’s Nikos Korfiatis discussing the issues arising from spreadsheet errors and the difficulty of detecting them. Nikos presented examples where spreadsheet errors went undetected and introduced some of the common pitfalls discussed from the European Spreadsheet Risk Interest Group (EuSpRiG) which is hosting its conference in July at Imperial College’s Data Science Institute. The standard practice of integrating spreadsheets in analytics pipelines has several issues related with reproducibility. Several new exciting opportunities for data scientists on generating unit tests using R, an open source programming language and software environment for statistical computing and graphics. R’s “markdown” and “testthat” packages for ensuring data quality in reporting are currently becoming mainstream and these developments could be discussed in a more practical forthcoming meetup session (datathon). The presentation concluded with a case study on how UK’s Government’s Digital Service (GDS), a part of the Cabinet Office, has used Rmarkdown in streamlining and simplifying its report publication pipeline.
The meetup event featured pitches from meetup participants. John Beer of Downham Market Development Ltd, who has strong ties with Novartis, provided an overview on his organizational activities involving Data Science and Analytics in Cancer Research and Pharma. Jason Young of Elevate Credit provided an overview of his company’s grow challenges and how difficult is to acquire talent in all aspects of the data science lifecycle. As John and Jason emphasized, more need to be done to address the data skills gap and Data Science Norfolk is a great avenue for more concentrated actions where business and academia can work together.
Finally, a round of applause was given from the audience to AVIVAs Maria Bean for kindly organizing the logistics of the event, and a tentative planning for an upcoming event in September was made.
Feature image photo credit – Tim Caynes