Proposal Description

Machine learning has a great and increasing importance in several branches of data analysis especially when using large data sets, given the very wide range of applications of machine learning tools. On the one hand, classical parametric modelling aims at providing interpretable results, while on the other, machine learning privileges a flexible approach in order to capture the nonlinear structure of the data.
The current debate focuses on advantages, limitations, and requirements of information extraction through machine learning, giving particular emphasis to explainability and interpretability of results and to the relationship with statistics.
This invited paper session is jointly proposed by the ISI Special Interest Group on Data Science and the International Society for Business Statistics.

Speakers:
Stefano Maria Iacus, European Centre for Algorithmic Transparency of the Joint Research Centre, European Commission, Title “Bridging Statistical Inference and Machine Learning: An Attention-Based Framework for Explainable Modeling”
Luis Ángel García-Escudero, Department of Statistics and Operational Research and IMUVA, University of Valladolid, Title” Choice of trimming proportion and number of clusters in robust clustering based on trimming”
Paola Vicard, Flaminia Musella, Università degli Studi Roma Tre, Daniela Marella, Title “Bayesian networks structural learning challenges for complex sampling design data”
Rahim Mahmoudvand, Bu-Ali Sina Universiy, Iran and University of Cagliari, Italy, Title “The Order of Things: Singular Spectrum Analysis as a Lens for Covariate-Driven Interpretability”