Abstract In the context of machine learning, classification is the procedure of predicting the class to which each element of a population belongs to. Most classification functions, for real world problems, are imperfect and thus require rigorous analysis for use in safety-critical applications such as health care. This paper proposes a software architecture for improving the trustworthiness and explainability of AI-based classifiers. The architecture combines a search-based approach with machine-learned explanations and satisfiability solving, to provide an indication of classification confidence and counterfactual explanation rules that are deductively verified to be consistent with the classifier. An implementation of the proposed architecture is evaluated on a medical case study of prognosis of Acute Coronary Syndrome (ACS). The evaluation shows that the proposed architecture is consistently able to complement each individual classification with an indication of confidence and an explanation, which is formally verified for consistency with the classifier. This contributes to foster trustworthy and explainable classification.

Abstract Artificial Intelligence systems could find many important applications in the medical field, holding excellent potential for improving disease diagnosis, treatment identification and selection. These opportunities are often jeopardized by the lack of interpretability of such systems, slowing down AI adoption. To overcome the issue, we first introduce an analytical framework exploiting multimodal deep learning for the classification of prostate lesions using Magnetic Resonance Imaging (MRI) data and clinical information on the patients. Then, we propose a multimodal explainability approach based on visual explanations to interpret the proposed model decision-making process and identify how the different modalities contribute to each specific prediction. Our findings, based on the PI-CAI Grand Challenge dataset, demonstrate the potential of combining multimodal data with eXplainable AI (XAI) to enhance prostate cancer diagnosis, improving model predictive performance, interpretability and understanding in treatment decision-making.

Federated Learning (FL) is a crucial approach for training large-scale AI models while preserving data locality, eliminating the need for centralised data storage. In collaborative learning settings, ensuring data quality is essential, and in FL, maintaining privacy requires limiting the knowledge accessible to the central orchestrator, which evaluates and manages client contributions. Accurately measuring and regulating the marginal impact of each client’s contribution needs specialised techniques. This work examines the relationship between one such technique—Shapley Values—and a client’s vulnerability to Membership inference attacks (MIAs). Such a correlation would suggest that the contribution index could reveal high-risk participants, potentially allowing a malicious orchestrator to identify and exploit the most vulnerable clients. Conversely, if no such relationship is found, it would indicate that contribution metrics do not inherently expose information exploitable for powerful privacy attacks. Our empirical analysis in a cross-silo FL setting demonstrates that leveraging contribution metrics in federated environments does not substantially amplify privacy risks.

Abstract This paper presents a methodology for integrating human expert knowledge into machine learning (ML) workflows to improve both model interpretability and the quality of explanations produced by explainable AI (XAI) techniques. We strive to enhance standard ML and XAI pipelines without modifying underlying algorithms, focusing instead on embedding domain knowledge at two stages: (1) during model development through expert-guided data structuring and feature engineering, and (2) during explanation generation via domain-aware synthetic neighbourhoods. Visual analytics is used to support experts in transforming raw data into semantically richer representations. We validate the methodology in two case studies: predicting COVID-19 incidence and classifying vessel movement patterns. The studies demonstrated improved alignment of models with expert reasoning and better quality of synthetic neighbourhoods. We also explore using large language models (LLMs) to assist experts in developing domain-compliant data generators. Our findings highlight both the benefits and limitations of existing XAI methods and point to a research direction for addressing these gaps.

Federated Learning (FL) is a widespread and well-adopted paradigm of decentralized learning that allows training one model from multiple sources without the need to directly transfer data between participating clients. Since its inception in 2015, it has been divided into numerous sub-fields that deal with application-specific issues, be it data heterogeneity or resource allocation. One such sub-field, Clustered Federated Learning (CFL), is dealing with the problem of clustering the population of clients into separate cohorts to deliver personalized models. Although few remarkable works have been published in this domain, the problem is still largely unexplored, as its basic assumption and settings are slightly different from standard FL. In this work, we present One-Shot Clustered Federated Learning (OCFL), a clustering-agnostic algorithm that can automatically detect the earliest suitable moment for clustering. Our algorithm is based on the computation of cosine similarity between gradients of the clients and a temperature measure that detects when the federated model starts to converge. We empirically evaluate our methodology by testing various one-shot clustering algorithms for over thirty different tasks on three benchmark datasets. Our experiments showcase the good performance of our approach when used to perform CFL in an automated manner without the need to adjust hyperparameters.

Abstract Human mobility data play a crucial role in understanding mobility patterns and developing analytical services across various domains such as urban planning, transportation, and public health. However, due to the sensitive nature of this data, accurately identifying privacy risks is essential before deciding to release it to the public. Recent work has proposed the use of machine learning models for predicting privacy risk on raw mobility trajectories and the use of shap for risk explanation. However, applying shap to mobility data results in explanations that are of limited use both for privacy experts and end-users. In this work, we present a novel version of the Expert privacy risk prediction and explanation framework specifically tailored for human mobility data. We leverage state-of-the-art algorithms in time series classification, as Rocket and InceptionTime, to improve risk prediction while reducing computation time. Additionally, we address two key issues with shap explanation on mobility data: first, we devise an entropy-based mask to efficiently compute shap values for privacy risk in mobility data; second, we develop a module for interactive analysis and visualization of shap values over a map, empowering users with an intuitive understanding of shap values and privacy risk.

Abstract The rise of sophisticated black-box machine learning models in Artificial Intelligence systems has prompted the need for explanation methods that reveal how these models work in an understandable way to users and decision makers. Unsurprisingly, the state-of-the-art exhibits currently a plethora of explainers providing many different types of explanations. With the aim of providing a compass for researchers and practitioners, this paper proposes a categorization of explanation methods from the perspective of the type of explanation they return, also considering the different input data formats. The paper accounts for the most representative explainers to date, also discussing similarities and discrepancies of returned explanations through their visual appearance. A companion website to the paper is provided as a continuous update to new explainers as they appear. Moreover, a subset of the most robust and widely adopted explainers, are benchmarked with respect to a repertoire of quantitative metrics.

eXplainable AI (XAI) involves two intertwined but separate challenges: the development of techniques to extract explanations from black-box AI models and the way such explanations are presented to users, i.e., the explanation user interface. Despite its importance, the second aspect has received limited attention so far in the literature. Effective AI explanation interfaces are fundamental for allowing human decision-makers to take advantage and oversee high-risk AI systems effectively. Following an iterative design approach, we present the first cycle of prototyping-testing-redesigning of an explainable AI technique and its explanation user interface for clinical Decision Support Systems (DSS). We first present an XAI technique that meets the technical requirements of the healthcare domain: sequential, ontology-linked patient data, and multi-label classification tasks. We demonstrate its applicability to explain a clinical DSS, and we design a first prototype of an explanation user interface. Next, we test such a prototype with healthcare providers and collect their feedback with a two-fold outcome: First, we obtain evidence that explanations increase users’ trust in the XAI system, and second, we obtain useful insights on the perceived deficiencies of their interaction with the system, so we can re-design a better, more human-centered explanation interface.

Explainable AI consists in developing mechanisms allowing for an interaction between decision systems and humans by making the decisions of the formers understandable. This is particularly important in sensitive contexts like in the medical domain. We propose a use case study, for skin lesion diagnosis, illustrating how it is possible to provide the practitioner with explanations on the decisions of a state of the art deep neural network classifier trained to characterize skin lesions from examples. Our framework consists of a trained classifier onto which an explanation module operates. The latter is able to offer the practitioner exemplars and counterexemplars for the classification diagnosis thus allowing the physician to interact with the automatic diagnosis system. The exemplars are generated via an adversarial autoencoder. We illustrate the behavior of the system on representative examples.

Increased availability of epidemiological data, novel digital data streams, and the rise of powerful machine learning approaches have generated a surge of research activity on real-time epidemic forecast systems. In this paper, we propose the use of a novel data source, namely retail market data to improve seasonal influenza forecasting. Specifically, we consider supermarket retail data as a proxy signal for influenza, through the identification of sentinel baskets, i.e., products bought together by a population of selected customers. We develop a nowcasting and forecasting framework that provides estimates for influenza incidence in Italy up to 4 weeks ahead. We make use of the Support Vector Regression (SVR) model to produce the predictions of seasonal flu incidence. Our predictions outperform both a baseline autoregressive model and a second baseline based on product purchases. The results show quantitatively the value of incorporating retail market data in forecasting models, acting as a proxy that can be used for the real-time analysis of epidemics.

In the last decade the usage and study of data mining and machine learning algorithms have received an increasing attention from several and heterogeneous fields of research. Learning how and why a certain algorithm returns a particular result, and understanding which are the main problems connected to its execution is a hot topic in the education of data mining methods. In order to support data mining beginners, students, teachers, and researchers we introduce a novel didactic environment. The Didactic Data Mining Environment (DDME) allows to execute a data mining algorithm on a dataset and to observe the algorithm behavior step by step to learn how and why a certain result is returned. DDME can be practically exploited by teachers and students for having a more interactive learning of data mining. Indeed, on top of the core didactic library, we designed a visual platform that allows online execution of experiments and the visualization of the algorithm steps. The visual platform abstracts the coding activity and makes available the execution of algorithms to non-technicians.

People living in highly populated cities increasingly experience decreased quality of life due to pollution and traffic congestion. With the objective of reducing the number of circulating vehicles, we investigate a novel approach to boost ride-sharing opportunities based on the knowledge of the human activities behind individual mobility demands. We observe that in many cases the activity motivating the use of a private car (e.g., going to a shopping mall) can be performed in many different places. Therefore, when there is the possibility of sharing a ride, people having a pro-environment behavior or interested in saving money can accept to fulfill their needs at an alternative destination. We thus propose activity-based ride matching (ABRM), an algorithm aimed at matching ride requests with ride offers, possibly reaching alternative destinations where the intended activity can be performed. By analyzing two large mobility datasets extracted from a popular social network, we show that our approach could largely impact urban mobility by resulting in an increase up to 54.69% of ride-sharing opportunities with respect to a traditional destination-oriented approach. Due to the high number of ride possibilities found by ABRM, we introduce and assess a subsequent ranking step to provide the user with the top-k most relevant rides only. We discuss how ABRM parameters affect the fraction of car rides that can be saved and how the ranking function can be tuned to enforce pro-environment behaviors.

Nowadays the analysis of diffusive phenomena occurring on top of complex networks represents a hot topic in the Social Network Analysis playground. In order to support students, teachers, developers and researchers in this work we introduce a novel simulation framework, NDlib. NDlib is designed to be a multi-level ecosystem that can be fruitfully used by different user segments. Upon the diffusion library, we designed a simulation server that allows remote execution of experiments and an online visualization tool that abstract the programmatic interface and makes available the simulation platform to non-technicians.

ComeWithMe is an activity oriented carpooling service that enlarges the candidate destinations of a ride request by considering alternative places where the desired activity can be performed. It is based on the observation that individuals often move towards a place to perform an activity while the activity is often not strictly associated with a single place, as one may go for shopping or eating to many different locations. Activity-oriented carpooling hugely increases the number of rides matching a query, thus introducing requirements on system responsiveness and ranking effectiveness that are not common to traditional carpooling services. The demoed system implements the ComeWithMe service in almost its entirety, and includes the back-end and a user-friendly mobile application for smart-phones aimed at achieving users' acceptance and usability.

Spatial events, such as lightning strikes or drops in moving vehicle speed, can be conceptualized as points in the space-time continuum. We consider real time monitoring scenarios in which the observer needs to detect significant (i.e., sufficiently big) spatio-temporal clusters of events as soon as they occur and track the further evolution of these clusters. Isolated spatial events and small clusters are of no interest (i.e., treated as noise) and should be hidden from the observer to avoid attention distraction and perceptual overload. The existing methods for stream clustering cannot enable on-the-fly separation of event clusters from the noise and immediate presentation of significant clusters and their evolution. We propose a novel algorithm tailored to this specific task and a visual analytics system that supports event stream monitoring by presenting detected event clusters and their evolution to the observer in real time.

The large availability of mobility data allows us to investigate complex phenomena about human movement. However this adundance of data comes with few information about the purpose of movement. In this work we address the issue of activity recognition by introducing Activity-Based Cascading (ABC) classification. Such approach departs completely from probabilistic approaches for two main reasons. First, it exploits a set of structural features extracted from the Individual Mobility Network (IMN), a model able to capture the salient aspects of individual mobility. Second, it uses a cascading classification as a way to tackle the highly skewed frequency of activity classes. We show that our approach outperforms existing state-of-the-art probabilistic methods. Since it reaches high precision, ABC classification represents a very reliable semantic amplifier for Big Data.

In recent years, the exponential growth of positioning-enabled devices have allowed us to study the mobility behavior of individuals analyzing their collected tracks. In this context, a small, but steadily increasing part of the literature is looking at the semantic aspects of mobility. This paper presents a contribution to this trend, and is concerned with the definition of semantic regularity profiles. We based our methodology on the entropy of both spatial and temporal frequency of visits of individuals to places to perform an activity. This allows us to define the concept of semantic regular or irregular user behavior identifying users who are more or less loyal to the same places in contrast to the flexibility in visiting different places to perform an activity. We experiment on a crowdsensed trajectory dataset annotated by the visited Points of Interest which represent the activity performed. Analysis evidence that the regularity depends on the particular activity to be performed.

This information about our GSM calls is stored by the TelCo operator in large volumes and with strict privacy constraints making it challenging the analysis of these fingerprints for inferring mobility behavior. This paper proposes a strategy for mobility behavior identification based on aggregated calling profiles of mobile phone users. This compact representation of the user call profiles is the input of the mining algorithm for automatically classifying various kinds of mobility behavior. A further advantage of having defined the call profiles is that the analysis phase is based on summarized privacy-preserving representation of the original data. We show how these call profiles permit to design a two step process - implemented into a system - based on a bootstrap phase and a running phase for classifying users into behavior categories. We evaluated the system in two case studies where individuals are classified into residents, commuters and visitors. We conclude the paper with a discussion which emphasizes the role of the call profiles for the design of a new collaboration model between data provider and data analyst.

In our market society, buyers are considered rational entities, driven by two utility functions: i) the amount of money spent, a universal quantity to be minimized; and ii) the individual needs to satisfy, a personal quantity, varying from person to person, to be maximized. In this paper, we propose an analytic framework based on big data to measure the personal utility function and we prove that this function has a stronger effect on customer behavior than the price. By focusing on the purchases in an Italian supermarket chain, we discover and describe a range effect of products: the more sophisticated the needs they satisfy, the more cost the customers are willing to pay to buy them, in terms of distance to travel more than in terms of the price of the item itself. We exhibit a striking empirical evidence of this theory by tracking the geographical information about points of sale and customers, in a large dataset containing tens of thousands of customers and thousands of products. We create a data mining framework able to scale to possibly hundreds of thousands, or millions, of customers and to let emerge from the data the knowledge about the actual range of each product. As an application of this finding, we show how it is possible to accurately predict how long a customer will travel (or which shop she will choose) to buy a product, as a function of the product's sophistication.

Place-oriented analysis of movement data, i.e., recorded tracks of moving objects, includes finding places of interest in which certain types of movement events occur repeatedly and investigating the temporal distribution of event occurrences in these places and, possibly, other characteristics of the places and links between them. For this class of problems, we propose a visual analytics procedure consisting of four major steps: 1) event extraction from trajectories; 2) extraction of relevant places based on event clustering; 3) spatiotemporal aggregation of events or trajectories; 4) analysis of the aggregated data. All steps can be fulfilled in a scalable way with respect to the amount of the data under analysis; therefore, the procedure is not limited by the size of the computer's RAM and can be applied to very large data sets. We demonstrate the use of the procedure by example of two real-world problems requiring analysis at different spatial scales.

The availability of massive network and mobility data from diverse domains has fostered the analysis of human behaviors and interactions. This data availability leads to challenges in the knowledge discovery community. Several different analyses have been performed on the traces of human trajectories, such as understanding the real borders of human mobility or mining social interactions derived from mobility and vice versa. However, the data quality of the digital traces of human mobility has a dramatic impact over the knowledge that it is possible to mine, and this issue has not been thoroughly tackled so far in literature. In this paper, we mine and analyze with complex network techniques a large dataset of human trajectories, a GPS dataset from more than 150k vehicles in Italy. We build a multi resolution grid and we map the trajectories with several complex networks, by connecting the different areas of our region of interest. Then we analyze the structural properties of these networks and the quality of the borders it is possible to infer from them. The result is a significant advancement in our understanding of the data transformation process that is needed to connect mobility with social network analysis and mining.