Rosa Arboretti Giancristofaro, rosa.arboretti@unipd.it

Research areas:

Design of Experiments and Nonparametric Inference
The ongoing research is focused on methodological solutions for the design and analysis of complex experiments for the comparison between different treatments (experimental conditions, processes, products) and the study of how input variables can affect one or more output variables (response variables). The complexity may be both in the characteristics of the experiment (block designs, repeated measures, restricted randomization, mixture designs), in the nature of multivariate or mixed response variables and in the kind of hypothesis tests (multivariate and directional alternatives, stochastic dominance). After applying Design of experiments (DoE) techniques, the nonparametric inference based on the nonparametric combination of dependent permutation tests (NPC), can allow to overcome most of the inferential problems. Developments in the context of NPC methodology can represent interesting solutions to support engineering research, industrial development and technological innovation.

Design of Experiments and Machine Learning techniques for Big Data Analytics
A further research topic concerns the combined use of Design of experiments (DoE) and Machine Learning (ML) techniques for the analysis of Big Data to make predictions and solve classification or product optimization problems. One of the criticism of ML techniques is the difficulty in identifying causal links between variables thus detecting only correlations. Furthermore, ML models tend to work as black boxes in which the chosen algorithm, after a training phase, independently proceeds in the analysis, making it difficult to understand. In order to face these problems, an innovative approach is under study to integrate the DoE methodology with ML modeling. Interesting applications are in the context of predictive maintenance for the development of fault detection systems, maintenance scheduling and resource optimization.

Keywords: Nonparametric tests, Big Data Analytics, Design of Experiments