CUSP LONDON SEMINAR SERIES In conjunction with the Human Centred Computing Research Group:
12 Dec 2019 – CUSP Centre, King’s College London – BH(S)5.01 – 2pm-3pm
Title: Mapping Human Population – A Data Science Approach
Where do people live? How do they commute? Where and when do they go on holiday? These are typical questions that are important for national statistical institutes. Whereas in the past, survey data and classical (frequentist) statistical inference were used to produce such statistics, nowadays, a variety of methods from Bayesian statistical inference and machine learning are applied, using a mix of administrative data and big data. Furthermore, data visualization is increasingly important to analyse the results and present them to the general public. In this presentation, I illustrate this new trend in official statistics with two research projects that are related to mapping human population.
Mobile phone network data can be used to estimate where people are during the day. In contrast to census data, which contains information about where people live, the ‘nighttime’ population, mobile phone network data opens up the possibility to estimate the ‘daytime’ population. A key challenge is to develop general methodology for using mobile phone network data. I will zoom in on one specific task, namely how to estimate the geographical location of a mobile phone, and present an Bayesian approach to solve this.
Today, statistics can be produced with much more spatial detail. An innovative visualization method that can be used to present spatial statistics is the dot map, in which every person is represented by a dot. I illustrate the dot map with some examples and describe the used methodology and software to create the dot map.
Biography: Martijn Tennekes has a Masters in knowledge engineering (now called data science), PhD in game theory (Maastricht University), currently working at Statistics Netherlands (‘the Dutch ONS’) as data scientist on data visualization, statistical programming, and the use of big data for official statistics.