Data Mining in Clinical Practice for the Quantification of Motor Impairment in Parkinson's Disease

Advances in biomedical signal acquisition systems for motion analysis have led to lowcost and ubiquitous wearable sensors which can be used to record movement data in different settings. This implies the potential availability of large amounts of quantitative data. It is then crucial to identify and to extract the information of clinical relevance from the large amount of available data. This quantitative and objective information can be an important aid for clinical decision making. Data mining is the process of discovering such information in databases through data processing, selection of informative data, and identification of relevant patterns. The databases considered in this thesis store motion data from wearable sensors (specifically accelerometers) and clinical information (clinical data, scores, tests). The main goal of this thesis is to develop data mining tools which can provide quantitative information to the clinician in the field of movement disorders. This thesis will focus on motor impairment in Parkinson's disease (PD). Different databases related to Parkinson subjects in different stages of the disease were considered for this thesis. Each database is characterized by the data recorded during a specific motor task performed by different groups of subjects. The data mining techniques that were used in this thesis are feature selection (a technique which was used to find relevant information and to discard useless or redundant data), classification, clustering, and regression. The aims were to identify high risk subjects for PD, characterize the differences between early PD subjects and healthy ones, characterize PD subtypes and automatically assess the severity of symptoms in the home setting.

Applicability of Process Mining Techniques in Business Environments

This thesis analyses problems related to the applicability, in business environments, of Process Mining tools and techniques. The first contribution is a presentation of the state of the art of Process Mining and a characterization of companies, in terms of their "process awareness". The work continues identifying circumstance where problems can emerge: data preparation; actual mining; and results interpretation. Other problems are the configuration of parameters by not-expert users and computational complexity. We concentrate on two possible scenarios: "batch" and "on-line" Process Mining. Concerning the batch Process Mining, we first investigated the data preparation problem and we proposed a solution for the identification of the "case-ids" whenever this field is not explicitly indicated. After that, we concentrated on problems at mining time and we propose the generalization of a well-known control-flow discovery algorithm in order to exploit non instantaneous events. The usage of interval-based recording leads to an important improvement of performance. Later on, we report our work on the parameters configuration for not-expert users. We present two approaches to select the "best" parameters configuration: one is completely autonomous; the other requires human interaction to navigate a hierarchy of candidate models. Concerning the data interpretation and results evaluation, we propose two metrics: a model-to-model and a model-to-log. Finally, we present an automatic approach for the extension of a control-flow model with social information, in order to simplify the analysis of these perspectives. The second part of this thesis deals with control-flow discovery algorithms in on-line settings. We propose a formal definition of the problem, and two baseline approaches. The actual mining algorithms proposed are two: the first is the adaptation, to the control-flow discovery problem, of a frequency counting algorithm; the second constitutes a framework of models which can be used for different kinds of streams (stationary versus evolving).

Digital Forensics AI: on Practicality, Optimality, and Interpretability of Digital Evidence Mining Techniques

Digital forensics as a field has progressed alongside technological advancements over the years, just as digital devices have gotten more robust and sophisticated. However, criminals and attackers have devised means for exploiting the vulnerabilities or sophistication of these devices to carry out malicious activities in unprecedented ways. Their belief is that electronic crimes can be committed without identities being revealed or trails being established. Several applications of artificial intelligence (AI) have demonstrated interesting and promising solutions to seemingly intractable societal challenges. This thesis aims to advance the concept of applying AI techniques in digital forensic investigation. Our approach involves experimenting with a complex case scenario in which suspects corresponded by e-mail and deleted, suspiciously, certain communications, presumably to conceal evidence. The purpose is to demonstrate the efficacy of Artificial Neural Networks (ANN) in learning and detecting communication patterns over time, and then predicting the possibility of missing communication(s) along with potential topics of discussion. To do this, we developed a novel approach and included other existing models. The accuracy of our results is evaluated, and their performance on previously unseen data is measured. Second, we proposed conceptualizing the term “Digital Forensics AI” (DFAI) to formalize the application of AI in digital forensics. The objective is to highlight the instruments that facilitate the best evidential outcomes and presentation mechanisms that are adaptable to the probabilistic output of AI models. Finally, we enhanced our notion in support of the application of AI in digital forensics by recommending methodologies and approaches for bridging trust gaps through the development of interpretable models that facilitate the admissibility of digital evidence in legal proceedings.

Learning Methods and Algorithms for Semantic Text Classification across Multiple Domains

Information is nowadays a key resource: machine learning and data mining techniques have been developed to extract high-level information from great amounts of data. As most data comes in form of unstructured text in natural languages, research on text mining is currently very active and dealing with practical problems. Among these, text categorization deals with the automatic organization of large quantities of documents in priorly defined taxonomies of topic categories, possibly arranged in large hierarchies. In commonly proposed machine learning approaches, classifiers are automatically trained from pre-labeled documents: they can perform very accurate classification, but often require a consistent training set and notable computational effort. Methods for cross-domain text categorization have been proposed, allowing to leverage a set of labeled documents of one domain to classify those of another one. Most methods use advanced statistical techniques, usually involving tuning of parameters. A first contribution presented here is a method based on nearest centroid classification, where profiles of categories are generated from the known domain and then iteratively adapted to the unknown one. Despite being conceptually simple and having easily tuned parameters, this method achieves state-of-the-art accuracy in most benchmark datasets with fast running times. A second, deeper contribution involves the design of a domain-independent model to distinguish the degree and type of relatedness between arbitrary documents and topics, inferred from the different types of semantic relationships between respective representative words, identified by specific search algorithms. The application of this model is tested on both flat and hierarchical text categorization, where it potentially allows the efficient addition of new categories during classification. Results show that classification accuracy still requires improvements, but models generated from one domain are shown to be effectively able to be reused in a different one.

Non Invasive Tools for Early Detection of Autism Spectrum Disorders

Autism Spectrum Disorders (ASDs) describe a set of neurodevelopmental disorders. ASD represents a significant public health problem. Currently, ASDs are not diagnosed before the 2nd year of life but an early identification of ASDs would be crucial as interventions are much more effective than specific therapies starting in later childhood. To this aim, cheap an contact-less automatic approaches recently aroused great clinical interest. Among them, the cry and the movements of the newborn, both involving the central nervous system, are proposed as possible indicators of neurological disorders. This PhD work is a first step towards solving this challenging problem. An integrated system is presented enabling the recording of audio (crying) and video (movements) data of the newborn, their automatic analysis with innovative techniques for the extraction of clinically relevant parameters and their classification with data mining techniques. New robust algorithms were developed for the selection of the voiced parts of the cry signal, the estimation of acoustic parameters based on the wavelet transform and the analysis of the infant’s general movements (GMs) through a new body model for segmentation and 2D reconstruction. In addition to a thorough literature review this thesis presents the state of the art on these topics that shows that no studies exist concerning normative ranges for newborn infant cry in the first 6 months of life nor the correlation between cry and movements. Through the new automatic methods a population of control infants (“low-risk”, LR) was compared to a group of “high-risk” (HR) infants, i.e. siblings of children already diagnosed with ASD. A subset of LR infants clinically diagnosed as newborns with Typical Development (TD) and one affected by ASD were compared. The results show that the selected acoustic parameters allow good differentiation between the two groups. This result provides new perspectives both diagnostic and therapeutic.

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.

Precision horticulture: application on apple orchards

Precision horticulture and spatial analysis applied to orchards are a growing and evolving part of precision agriculture technology. The aim of this discipline is to reduce production costs by monitoring and analysing orchard-derived information to improve crop performance in an environmentally sound manner. Georeferencing and geostatistical analysis coupled to point-specific data mining allow to devise and implement management decisions tailored within the single orchard. Potential applications range from the opportunity to verify in real time along the season the effectiveness of cultural practices to achieve the production targets in terms of fruit size, number, yield and, in a near future, fruit quality traits. These data will impact not only the pre-harvest but their effect will extend to the post-harvest sector of the fruit chain. Chapter 1 provides an updated overview on precision horticulture , while in Chapter 2 a preliminary spatial statistic analysis of the variability in apple orchards is provided before and after manual thinning; an interpretation of this variability and how it can be managed to maximize orchard performance is offered. Then in Chapter 3 a stratification of spatial data into management classes to interpret and manage spatial variation on the orchard is undertaken. An inverse model approach is also applied to verify whether the crop production explains environmental variation. In Chapter 4 an integration of the techniques adopted before is presented. A new key for reading the information gathered within the field is offered. The overall goal of this Dissertation was to probe into the feasibility, the desirability and the effectiveness of a precision approach to fruit growing, following the lines of other areas of agriculture that already adopt this management tool. As existing applications of precision horticulture already had shown, crop specificity is an important factor to be accounted for. This work focused on apple because of its importance in the area where the work was carried out, and worldwide.

Dynamics of coastal aquifers: data-driven forecasting and risk analysis

This work is focused on the study of saltwater intrusion in coastal aquifers, and in particular on the realization of conceptual schemes to evaluate the risk associated with it. Saltwater intrusion depends on different natural and anthropic factors, both presenting a strong aleatory behaviour, that should be considered for an optimal management of the territory and water resources. Given the uncertainty of problem parameters, the risk associated with salinization needs to be cast in a probabilistic framework. On the basis of a widely adopted sharp interface formulation, key hydrogeological problem parameters are modeled as random variables, and global sensitivity analysis is used to determine their influence on the position of saltwater interface. The analyses presented in this work rely on an efficient model reduction technique, based on Polynomial Chaos Expansion, able to combine the best description of the model without great computational burden. When the assumptions of classical analytical models are not respected, and this occurs several times in the applications to real cases of study, as in the area analyzed in the present work, one can adopt data-driven techniques, based on the analysis of the data characterizing the system under study. It follows that a model can be defined on the basis of connections between the system state variables, with only a limited number of assumptions about the "physical" behaviour of the system.

Learning representations for graph-structured socio-technical systems

The recent widespread use of social media platforms and web services has led to a vast amount of behavioral data that can be used to model socio-technical systems. A significant part of this data can be represented as graphs or networks, which have become the prevalent mathematical framework for studying the structure and the dynamics of complex interacting systems. However, analyzing and understanding these data presents new challenges due to their increasing complexity and diversity. For instance, the characterization of real-world networks includes the need of accounting for their temporal dimension, together with incorporating higher-order interactions beyond the traditional pairwise formalism. The ongoing growth of AI has led to the integration of traditional graph mining techniques with representation learning and low-dimensional embeddings of networks to address current challenges. These methods capture the underlying similarities and geometry of graph-shaped data, generating latent representations that enable the resolution of various tasks, such as link prediction, node classification, and graph clustering. As these techniques gain popularity, there is even a growing concern about their responsible use. In particular, there has been an increased emphasis on addressing the limitations of interpretability in graph representation learning. This thesis contributes to the advancement of knowledge in the field of graph representation learning and has potential applications in a wide range of complex systems domains. We initially focus on forecasting problems related to face-to-face contact networks with time-varying graph embeddings. Then, we study hyperedge prediction and reconstruction with simplicial complex embeddings. Finally, we analyze the problem of interpreting latent dimensions in node embeddings for graphs. The proposed models are extensively evaluated in multiple experimental settings and the results demonstrate their effectiveness and reliability, achieving state-of-the-art performances and providing valuable insights into the properties of the learned representations.