Functional and Object-oriented Data Analysis for Mobility

Statisticians are currently asked to develop tools to analyze extremely complex data often far from the standard Euclidean framework (e.g., curves, networks, compositions, images, tensors). The Statistics group at MOX proposes a paradigm of statistical investigation based on the idea of developing sound approaches to data analysis able to provide valuable and trustful results. Merging Nonparametric Inference & Prediction with Object-oriented Data Analysis is the way this task is pursued starting from functional data (the most investigated case of object-data) and then move toward Hilbert data and purely metric object-data for which a meaningful Hilbertian embedding cannot be found.

Mobility of people and goods is a natural setting where these types of data are often encountered: mobility in a specific area is often described as a time-varying directional weighted networks, flows between origins and destination as georeferenced functions, the mix of different types of vehicles in different countries as compositional data, charging stations for electric vehicle as described stochastic matrices describing the occupancy dynamics, dense urban textures as a spatial fields of tensors describing the local directionality of roads.

The newly developed models and algorithms have been already successfully used in many funded projects based on an active partnership between the MOX lab, other domain-specific research groups at Politecnico di Milano, private companies and national or international institutions. Two recent examples are the project Safari Njema and the project Designing a National Network Supporting Electric Mobility.

Safari njema: From informal mobility to mobility policies through big data analysis (March 2019 – June 2020) is part of the CSR program of Politecnico di Milano. The project explores ​bottom-up reorganizational place based replicable solutions alternative and complementary to traditional public policies and huge infrastructural investments to reduce mobility poverty in Sub-Saharan countries. Indeed, 80% of everyday mobility in African cities is supported by informal inefficient and unsafe mobility systems that hinder economic and social development.  The pilot case is taking place in the urban area of Greater Maputo (​Mozambique​​). Partners from the private sectors are Cuebiq (a world-leading company in the collection of geo-referenced data) and Vodacom (one of the largest mobile network providers in Mozambique). Institutional partners of the project are Unversidade Eduardo Mondlane and the Italian Agency for Development and Cooperation. The project is strongly multidisciplinary with five departments of Politecnico di Milano being involved in the project (i.e., Dept of Mathematics, Dept of Architecture and Urban Studies, Dept of Management, Dept of Computer Sciences, and Dept of Design).


Anonymized aggregated GPS location of mobile phones in the area of Greater Maputo. Data source: Cuebiq.

Designing a National Network Supporting Electric Mobility (April 2016 – June 2016). The project aims at designing a national network of charging stations for electric vehicles which could trigger a mass diffusion of this type of transport modality. Starting from a data fusion approach (based on the integration between census data, road sensor data, mobility surveys, and an analysis of macro-economic trends) the project provides the number and type of charging stations required in each Italian municipality, major road, and high way. An analysis of the economic and environmental impacts is also performed. The project is in partnership with Enel Foundation (Enel is the major energy supplier in Italy) and the Dept of Mathematics, the Dept of Management, and the Dept of Energy being part of the projects.



Time-varying bike-sharing mobility networks between different neighborhoods in the city of Milan. Data source: Clear Channel.


Diffusion tensors of bike trips in the historical center of Milan. Data source: Clear Channel.