Knowledge transfer and Collaboration
With the aim of extending its impact beyond the academic field, the Data Science Lab conducts various applied research projects. Learn about our most recent projects.
Economy and Society
Abrazar: Early detection of physical violence against children at the household level in Argentina through the application of predictive models.
Observational Study of User Behavior on the Road: Development of an efficient sampling method that allows determining with national and regional representativeness the presence of distraction factors and the rate of use of road safety elements (seat belt, helmet, child restraint systems, daytime running lights) by drivers and occupants of vehicles with two or more wheels.
Online Auctions: How to sell products with zero marginal cost and infinite stock? Development of a mathematical model applicable to real sales, with a distinctive feature: the community interested in the product sets the price through a competitive auction mechanism. Behavioral experiments show that the model generates higher profits than traditional fixed-price methods.
NILUS: Development of spatial statistics methodology to identify food deserts.
ADIM: Alternative rating measurement project. Collaboration with methodology development and subsequent analysis of longitudinal and demographic data to study viewers behavior in the new paradigm of multiple screens at home. The aim was determining the viewers presence, attention and use of the mobile phone during the broadcast content versus those in which there are commercial advertisements.
Smartwatch: Algorithm development, using smartwatch accelerometer data, to measure hand position during cardiopulmonary resuscitation.
Natural Sciences | Biology. Geology. Neurosciences.
Microair Polar: Methodologies development to interpret and classify wind paths using clustering techniques, functional data and depths.
Analysis of guanaco footprints and locomotion as an analogous model for extinct native ungulates. Construction of locomotion metrics among different steps and methods for comparison with the fossil record.
Microair Polar 2: Development of a new machine learning procedure to characterize central and outlier curves in a set of trajectories. An efficient algorithm is proposed to work with large data sets. We apply this new technique to identify the main routes followed by air masses transporting microorganisms to Byers Peninsula (Livingston Island, Antarctica).
Neuromat: Grouping according to distributions for Electroencephalogram (EEG) data. Statistical learning tools development to identify models used by the brain to perform compression tasks. The goal was to propose an unsupervised learning procedure based on the idea of grouping data by the probabilistic law that generated them.
Sample clustering strategies: How to use few reagents and diagnose a very large population? Sample pooling strategies involve combining samples from many individuals into a single reagent. Considering PCR technology sensitivity and the viral load concentration of infected individuals, we have been able to provide grouping strategies minimizing the number of tests to be performed.