Our colleagues Vlad Alexe and Norbert Petrovici will present tomorrow at the CEBSS Winter Session 2026, an edition of scientific seminars dedicated to the socio-human sciences, a study aimed at reconstructing daily mobility in Cluj County.
Date: 22.01.26, 2 pm, room 118, FSEGA headquarters
or online at Zoom link: https://us02web.zoom.us/j/83387907164 , Zoom Meeting ID: 833 8790 7164
Short abstract: This study uses passive mobile phone data to study daily mobility in Cluj County. Individual trajectories are reconstructed with a constrained Hidden Markov Model. Aggregated commuting flows are then analysed using spatial panel models. The results show a fragmented metropolitan structure and strong spatial dependence between residential areas and employment locations.
Long abstract: Passive mobile phone data from LTE networks make it possible to observe daily mobility at large scale, with high time resolution and broad territorial coverage. Although the data do not record true geographic positions, they allow the reconstruction of approximate individual trajectories and the construction of origin–destination matrices. These outputs describe commuting patterns, daily rhythms, and functional links between urban, suburban, and rural areas. The study has two objectives. The first is to reconstruct daily mobility trajectories for October 2023 in Cluj County. The second is to model aggregated home–work commuting flows using a spatial panel framework that combines mobility data with territorial and employment information.
The main technical challenge of the first objective lies in the indirect nature of the observations. The data consist of time-ordered connections between anonymised devices and mobile phone antennas, not of real locations. Standard geometric delineations of antenna coverage introduce structural uncertainty, which leads to location errors and unstable aggregated flows. To address this issue, we rely on operator-grade best-server radio maps at 25-metre resolution. These maps are combined with signalling sequences in a constrained Hidden Markov Model. The model enforces physical limits on movement, stabilises location sequences, and assigns a confidence score to each step. Applied to one month of data in Cluj County, the method produces coherent daily trajectories and hourly origin–destination matrices. Clear differences emerge between weekdays and weekends. The validation of the algorithm is carried out using mobile phone records that include both GPS traces and antenna connections. The results show a strong match at the scale of radio coverage geometry, with larger deviations only in areas affected by difficult signal propagation.
The second objective analyses daily commuting using aggregated trajectories within a fixed-effects spatial panel regression framework built on daily origin–destination matrices. The dependent variable captures commuter inflows into each spatial unit. Explanatory variables include sectoral employment structure, ownership type of employing firms, and residential location characteristics. Spatial error models account for unobserved territorial dependence linked to infrastructure, land use, and access constraints.
The results point to a fragmented metropolitan system. Jobs concentrate in a limited number of functional nodes, while housing spreads across suburban rings and rural areas. This pattern strengthens spatial dependence between places of residence and places of work. In this setting, daily mobility acts as a tracer of urban change, revealing where daily coordination stabilises and where territorial gaps in distance, travel time, and access to jobs continue to grow.
