📅 Use of non-traditional data sources to nowcast migration trends through Artificial Intelligence technologies
Conference organised by University of Oxford’s Migration and Mobility Network (MMN) and Nuffield College, University of Oxford.
Talk included in “Session 1a: How do we measure migration? Methods and advancements”.
Abstract
In the last years many researchers have proposed the use of non-traditional data sources to study migration, including so-called social Big Data such as online social networks. Many types of data exist, still very scattered and heterogeneous: in the variety of this context, integration is not straightforward. In general, these works have been performed by Computer Scientists, since they require special skills to meaningfully embed and combine traditional and novel datasets. Our work describes the use of alternative types of data, a new multi-feature dataset and a new indicator that could significantly contribute to the study of migration and to forecast emerging trends through the use of Artificial Intelligence technologies. This approach is intended to find an alternative methodology to ultimately answer open questions about the human mobility framework (i.e. nowcasting flows and stocks, studying integration of multiple sources and knowledge, and investigating migration drivers).
For this purpose we provide the Multi-aspect Integrated Migration Indicators (MIMI) dataset that we built by integrating official data about bidirectional human mobility (i.e. traditional flow and stock data, gathered from sources like Eurostat and United Nations) with multidisciplinary features and original indicators, including the Facebook Social Connectedness Index (SCI). This latter measures the relative probability that two individuals across two countries are friends with each other on Facebook: it guarantees an ethical and anonymized collection of information on users and their friendships. The inclusion of this indicator in the dataset enables it to be exploited as a non-traditional way to describe, understand and nowcast international migration. Then we introduced a new measure, likewise included in the dataset, consisting in a formulation built to mirror the mathematical structure of the SCI: the Bidirectional Migration Probability (BMP) index, which takes into account both the inflows and outflows shared by two countries, and measures the relative probability of a person to be a migrant from country i to j and vice versa. BMP indicator allows to portray and predict bilateral migration trends relying on the intensity of social networking, since it shows significant correlations with SCI.
We believe SCI and our integrated dataset can be employed to study migration drivers, along with other traditionally used measures (cultural differences indicators, GDP, distance between countries, etc), through Machine Learning techniques, so that to link and combine the statistical and computational study of migration phenomenon with interdisciplinary perspectives (geographical, demographic, economic, sociological, anthropological). Indeed, the knowledge combined in the dataset is designed to develop a ML model able to extract novel information, analyze patterns and, from the strength of Facebook connectivity between countries, nowcast and forecast both present and future bilateral migration trends. The long-term perspective of this research is to build trustworthy and reliable predictions for future changing by using new ways of measuring and characterizing international migration, and by using advanced technologies such as Artificial Neural Networks.
Audience size and countries addressed. Over 120 people registered for in-person activities and over 200 people online, having affiliation in the following countries: Italy, U.S., GB, Brazil, Mexico, Spain, Poland, Germany, Argentina, Peru, Nepal, Algeria, France, Netherlands, Nigeria, Turkey, Sri Lanka, Australia, Belgium.
