Meaningful insights: explainability techniques for black-box models on tabular data

Artificial Intelligence (AI) and Machine Learning (ML) are novel data analysis techniques providing very accurate prediction results. They are widely adopted in a variety of industries to improve efficiency and decision-making, but they are also being used to develop intelligent systems. Their success grounds upon complex mathematical models, whose decisions and rationale are usually difficult to comprehend for human users to the point of being dubbed as black-boxes. This is particularly relevant in sensitive and highly regulated domains. To mitigate and possibly solve this issue, the Explainable AI (XAI) field became prominent in recent years. XAI consists of models and techniques to enable understanding of the intricated patterns discovered by black-box models. In this thesis, we consider model-agnostic XAI techniques, which can be applied to Tabular data, with a particular focus on the Credit Scoring domain. Special attention is dedicated to the LIME framework, for which we propose several modifications to the vanilla algorithm, in particular: a pair of complementary Stability Indices that accurately measure LIME stability, and the OptiLIME policy which helps the practitioner finding the proper balance among explanations' stability and reliability. We subsequently put forward GLEAMS a model-agnostic surrogate interpretable model which requires to be trained only once, while providing both Local and Global explanations of the black-box model. GLEAMS produces feature attributions and what-if scenarios, from both dataset and model perspective. Eventually, we argue that synthetic data are an emerging trend in AI, being more and more used to train complex models instead of original data. To be able to explain the outcomes of such models, we must guarantee that synthetic data are reliable enough to be able to translate their explanations to real-world individuals. To this end we propose DAISYnt, a suite of tests to measure synthetic tabular data quality and privacy.

I beni relazionali nella ricerca sociologica: che cosa sono e quali effetti producono

This paper studies relational goods as immaterial assets creating real effects in society. The work starts answering to this question: what kind of effects do relational goods produce? After an accurate literature examination we suppose relational goods are social relations of second order. In the hypotesis they come from the emergence of two distinct social relations: interpersonal and reflexive relations. We describe empirical evidences of these emergent assets in social life and we test the effects they produce with a model. In the work we focus on four targets. First of all we describe the emergence of relational goods through a mathematical model. Then we individualize social realities where relational goods show evident effects and we outline our scientific hypotesis. The following step consists in the formulation of empirical tests. At last we explain final results. Our aim is to set apart the constitutive structure of relational goods into a checkable model coherently with the empirical evidences shown in the research. In the study we use multi-variate analysis techniques to see relational goods in a new way and we use qualitative and quantitative strategies. Relational goods are analysed both as dependent and independent variable in order to consider causative factors acting in a black-box model. Moreover we analyse effects of relational goods inside social spheres, especially in third sector and capitalistic economy. Finally we attain to effective indexes of relational goods in order to compare them with some performance indexes.

Analisi della dinamica e composizione chimica del particolato atmosferico mediante un modello alla mesoscala

The vertical profile of aerosol in the planetary boundary layer of the Milan urban area is studied in terms of its development and chemical composition in a high-resolution modelling framework. The period of study spans a week in summer of 2007 (12-18 July), when continuous LIDAR measurements and a limited set of balloon profiles were collected in the frame of the ASI/QUITSAT project. LIDAR observations show a diurnal development of an aerosol plume that lifts early morning surface emissions to the top of the boundary layer, reaching maximum concentration around midday. Mountain breeze from Alps clean the bottom of the aerosol layer, typically leaving a residual layer at around 1500-2000 m which may survive for several days. During the last two days under analysis, a dust layer transported from Sahara reaches the upper layers of Milan area and affects the aerosol vertical distribution in the boundary layer. Simulation from the MM5/CHIMERE modelling system, carried out at 1 km horizontal resolution, qualitatively reproduced the general features of the Milan aerosol layer observed with LIDAR, including the rise and fall of the aersol plume, the residual layer in altitude and the Saharan dust event. The simulation highlighted the importance of nitrates and secondary organics in its composition. Several sensitivity tests showed that main driving factors leading to the dominance of nitrates in the plume are temperature and gas absorption process. A modelling study turn to the analysis of the vertical aerosol profiles distribution and knowledge of the characterization of the PM at a site near the city of Milan is performed using a model system composed by a meteorological model MM5 (V3-6), the mesoscale model from PSU/NCAR and a Chemical Transport Model (CTM) CHIMERE to simulate the vertical aerosol profile. LiDAR continuous observations and balloon profiles collected during two intensive campaigns in summer 2007 and in winter 2008 in the frame of the ASI/QUITSAT project have been used to perform comparisons in order to evaluate the ability of the aerosol chemistry transport model CHIMERE to simulate the aerosols dynamics and compositions in this area. The comparisons of model aerosols with measurements are carried out over a full time period between 12 July 2007 and 18 July 2007. The comparisons demonstrate the ability of the model to reproduce correctly the aerosol vertical distributions and their temporal variability. As detected by the LiDAR, the model during the period considered, predicts a diurnal development of a plume during the morning and a clearing during the afternoon, typically the plume reaches the top of the boundary layer around mid day, in this time CHIMERE produces highest concentrations in the upper levels as detected by LiDAR. The model, moreover can reproduce LiDAR observes enhancement aerosols concentrations above the boundary layer, attributing the phenomena to dust out intrusion. Another important information from the model analysis regard the composition , it predicts that a large part of the plume is composed by nitrate, in particular during 13 and 16 July 2007 , pointing to the model tendency to overestimates the nitrous component in the particular matter vertical structure . Sensitivity study carried out in this work show that there are a combination of different factor which determine the major nitrous composition of the “plume” observed and in particular humidity temperature and the absorption phenomena are the mainly candidate to explain the principal difference in composition simulated in the period object of this study , in particular , the CHIMERE model seems to be mostly sensitive to the absorption process.