Chair: Masci
Ragni, A., Cavinato, L., Vantini, S., Ieva, F. Assessment of the effect of environmental exposure on Heart Rate through Interrupted Time Series on walk-about data in Milan
Abstract
The lifelong environmental exposures has shown to have a significant impact on human health, contributing to conditions such as respiratory diseases, cardiovascular issues, and cancer. Traditional monitoring methods remain rough and lack a proper spatiotemporal resolution. This study demonstrates the potential of wearable devices to bridge this gap by providing real-time data on environmental exposure and physiological responses (e.g., heart rate). Using an interrupted time series model, we analyze fatigue recovery by modelling heart rate variations in response to different urban microenvironments, each associated with distinct levels of exposure to environmental pollutants, offering preliminary insights into the physiological impacts of pollution exposure. As a result, we underscore the potential of wearable technologies in improving health monitoring and life quality.
Palummo, A., Arnone, E., Sangalli, L. M. Physics-informed nonparametric regression of largescale space-time datasets
Abstract
In this work, we introduce a novel approach within the physics-informed statistical learning framework for penalized nonparametric estimation of smooth space time fields on irregular geometries. Our method leverages an innovative iterative numerical discretization technique that significantly accelerates computation while maintaining accuracy. This enhanced efficiency makes our approach well-suited for complex and large-scale data analysis tasks, at population-level, such as functional principal component analysis. To illustrate its utility, we apply the presented methodology to the analysis of space-dependent spectra localized within the intricated anatomical structure of the brain volume.
Scielzo-Ortiz, F., Grané, A., Albarrán, I. Robust clustering for large datasets of weighted and mixed type data. An application to mental well-being in older adults
Abstract
The main objective of this paper is to create profiles of older adults to better understand differing levels of emotional well-being across Europe. Data comes from the latest wave of the Survey of Health, Ageing and Retirement in Europe, carried out in 26 countries and representing over 181 million aged individuals in Europe. By using the information of around 80 variables of mixed type, we design composite indicators focused on quality of life in older age, loneliness, social integration and social connectedness. In a second step, indicators are combined with a collection of descriptive variables by computing pairwise robust generalized Gower distance for weighted data. Finally, Fast k-medoids clustering algorithm is applied to obtain the profiles. Pyhton packages PyDistances and FastKmedoids are used for the computations.
Ouedrhiri, O., Banouar, O., Faghihi, U., Toure, F. Quantum Distance Measures for Enhanced Clustering Algorithms
Abstract
Clustering algorithms (The K-means and the Fuzzy C-means) play a crucial role in discovering hidden patterns in datasets. Integrating quantum distance metrics to assign each data point to a certain cluster can enhance them and reduce their time complexity. This paper studies the use of the quantum trace distance, the Hellinger distance, and the fidelity measure in clustering algorithms. We compare the time complexity and performance with the classical versions. The experimentation is done on three different datasets: Iris, Wine and Breast Cancer. The partitioning quality is measured using the Silhouette, Davies-Bouldin Score , the Partition Coefficient (FPC) and Partition Entropy (FPE). The results prove that using quantum distance measures can lead to efficient and faster clustering algorithms.