General efficient crop protection in olive orchards is challenged by the need for precise agrochemical application minimizing environmental impact while ensuring effective spray coverage. Conventional approaches often lead to poor spray distribution, heavy off-target losses and a lack of real-time adaptability in variable field conditions. Such findings imply the need for an approach or system that can adjust application parameters dynamically to optimize both effectiveness and sustainability.This paper introduces an advanced AI-driven architecture designed to meet these challenges by integrating Internet of Things (IoT), machine learning, and digital twin technologies. The core of this approach is the development of a Digital Tree, a virtual model of olive trees that accurately simulates and predicts spray distribution and interactions with environmental variables. IoT sensors in the field collect real-time data on spray behavior and field conditions, which machine learning algorithms then process to refine application parameters dynamically. By enabling data-driven, adaptive decision-making, the digital twin supports optimal spray distribution, reduces off-target impact, and enhances environmental sustainability. This integrated solution offers a scalable and replicable methodology for smart agriculture, advancing precision and sustainability in complex agricultural environments.

Expert recommendation is a central task in question-answering communities that involves facing several issues, such as quantifying users' expertise and replying propensity, discounting both over time, and taking advantage of the social graph of cooperative asking-answering interactions. In this paper, we propose two model-based approaches to recommend repliers in question-answering communities, whose novelty lies in the above issues being tackled jointly. The devised approaches focus on tags to avoid processing large amounts of (long and short) text messages.The first approach routes questions to answerers with the greatest expertise in the topics marked by question tags at routing time. To this end, the expertise of community members under such tags is distinguished between shown and unknown. More precisely, the expertise shown by answerers in previously-covered question topics is discounted by looking at the temporal information, votes, and tags of their past answers. Instead, the discounted degree of the unknown expertise of answerers in not-yet-covered question topics is predicted through the embeddings of tags and answerers into a space of latent factors. Such embeddings stem from Gibbs sampling inference under a fully-Bayesian latent-factor model of the temporally-discounted tag-based expertise of users and the social graph of their replying behavior.The second approach recommends repliers by also accounting for their temporally-discounted, tag-based answering propensity. In essence, a higher propensity to answer questions with certain tags is more likely to be expected in those community members who have answered a larger number of earlier questions with those tags, especially in the recent past.An extensive experimentation with real-world benchmark datasets demonstrates the superiority of our approaches over state-of-the-art competitors in recommendation effectiveness, according to several evaluation criteria. In particular, the average competitive gain of the first and second approaches is 11% and 13%, respectively, across the datasets.

The behavioral analysis of smart devices plays a key role in enforcing security for IoT environments. In particular, anomalous patterns can be targeted in the behavior of smart devices as potential IoT cybersecurity threats. In this article, an explainable machine-learning approach is proposed for dealing with behavioral anomalies. Essentially, a rule-based classifier is inferred from the observed behavior of smart devices, to detect and explain patterns of behavioral anomalies. Predictive association modeling is adopted in the formulation of the classifier, to achieve superior effectiveness in detecting behavioral patterns and ensuring clear explanations of both these latter and their classifications. Moreover, the specifically-conceived design of the classifier reduces the number of tunable parameters to one. An extensive empirical evaluation is comparatively carried out on real-world benchmark data. The experimental results reveal the effectiveness, robustness, and scalability of the proposed approach.

We study the seamless integration of community discovery and behavioral role analysis, in the domain of networks with node attributes. In particular, we focus on unifying the two tasks, by explicitly harnessing node attributes and behavioral role patterns in a principled manner. To this end, we propose two Bayesian probabilistic generative models of networks, whose novelty consists in the interrelationship of overlapping communities, roles, their behavioral patterns and node attributes. The devised models allow for a variety of exploratory, descriptive and predictive tasks. These are carried out through mean-field variational inference, which is in turn mathematically derived and implemented into a coordinate-ascent algorithm. A wide spectrum of experiments is designed, to validate the devised models against three classes of state-of-the-art competitors using various real-world benchmark data sets from different social networking services.

We propose a new statistical-learning approach to marrying topic modeling and document clustering. In particular, a Bayesian generative model of text collections is developed, in which the two foresaid tasks are incorporated as coupled latent factors, that govern document wording. The latter consists of word embeddings, so as to capture the semantic and syntactic regularities among words. Collapsed Gibbs sampling is derived mathematically and implemented algorithmically, along with parameter estimation, with the aim to jointly perform topic modeling and document clustering through Bayesian reasoning. Comparative tests on benchmark real-world corpora reveal the effectiveness of the devised approach in clustering collections of text documents and coherently recovering their semantics.

Document clustering and topic modeling are fundamental tasks in text mining, that can be unified to reciprocally enhance each other. In this paper, we present a machine learning approach to the joint modeling and interdependent fulfilment of both tasks. In particular, document clustering and topic modeling are seamlessly interrelated under an innovative Bayesian generative model of clusters, topics and contents in text corpora. Such a model assumes that text corpora result from a generative process, in which clusters and topics act as connected latent factors. Essentially, clusters are initially associated with descriptive and actionable topic distributions, that enforce cluster coherence. The individual documents are then assigned to one respective cluster and worded accordingly. Under the devised model, document clustering and topic modeling can be simultaneously performed in an interdependent manner simply by Bayesian reasoning. For this purpose, the mathematical details regarding collapsed Gibbs sampling as well as parameter estimation are derived and implemented into an approximate inference algorithm. Comparative experiments on standard benchmark text corpora reveal the effectiveness of our approach at jointly clustering text documents and unveiling their semantics in terms of coherent topics.

We propose an approach to clustering XML-based corpora of healthcare documents by their latent topic similarity. Our approach is a two-step process. Initially, the latent topic distributions of the input healthcare documents are inferred, by performing collapsed Gibbs sampling and parameter estimation under an XML topic model. Subsequently, the inferred distributions are grouped through established clustering techniques.

Community discovery and role assignment have been recently integrated into an unsupervised approach for the exploratory analysis of overlapping communities and inner roles in networks. However, the formation of ties in these prototypical research efforts is not truly realistic, since it does not account for a fundamental aspect of link establishment in real-world networks, i.e., the explicative reasons that cause interactions among nodes. Such reasons can be interpreted as generic requirements of nodes, that are met by other nodes and essentially pertain both to the nodes themselves and to their interaction contexts (i.e., the respective communities and roles). In this article, we present two new model-based machine-learning approaches, wherein community discovery and role assignment are seamlessly integrated and simultaneously performed through approximate posterior inference in Bayesian mixed-membership models of directed networks. The devised models account for the explicative reasons governing link establishment in terms of node-specific and contextual latent interaction factors. The former are inherently characteristic of nodes, while the latter are characterizations of nodes in the context of the individual communities and roles. The generative process of both models assigns nodes to communities with respective roles and connects them through directed links, which are probabilistically governed by their node-specific and contextual interaction factors. The difference between the proposed models lies in the exploitation of the contextual interaction factors. More precisely, in one model, the contextual interaction factors have the same impact on link generation. In the other model, the contextual interaction factors are weighted by the extent of involvement of the linked nodes in the respective communities and roles. We develop MCMC algorithms implementing approximate posterior inference and parameter estimation within our models. Finally, we conduct an intensive comparative experimentation, which demonstrates their superiority in community compactness and link prediction on various real-world and synthetic networks.

Recommendation on signed social rating networks is studied through an innovative approach. Bayesian probabilistic modeling is used to postulate a realistic generative process, wherein user and item interactions are explained by latent factors, whose relevance varies within the underlying network organization into user communities and item groups. Approximate posterior inference captures distrust propagation and drives Gibbs sampling to allow rating and (dis)trust prediction for recommendation along with the unsupervised exploratory analysis of network organization. Comparative experiments reveal the superiority of our approach in rating and link prediction on Epinions and Ciao , besides community quality and recommendation sensitivity to network organization.

We study XML partitioning via unsupervised topic modeling. A new mixed-membership Bayesian generative model of the latent topics in XML corpora is proposed. Approximate posterior inference and parameter estimation are derived for the devised XML topic model and implemented by a Gibbs sampling algorithm. This is used to infer the topic distributions of the input XML documents. In turn, such distributions are separated to divide the whole XML corpus by latent-topic similarity. Experiments on real-world XML corpora reveal an overcoming effectiveness with respect to several state-of-the-art competitors.

Conventional approaches to XML clustering by content and structure are generally affected by a limitation due to the adoption of the bag-of-word model for the representation of their textual contents. This choice may lead to consider structure-constrained textual items of separate XML documents as related, even though the actual meaning of such items in their respective contexts is different. To overcome such a limitation, we propose XML clustering by structure-constrained phrases. The latter is a previously unexplored method relying on the more accurate bag-of-phrase model of the XML textual content, with which to better preserve the meaning of the structure-constrained content items for improved clustering effectiveness. In order to conduct an in-depth and systematic study of the effectiveness of the proposed method, we develop a parameter-free prototypical approach to XML partitioning, which projects the XML documents into a space of XML features representing fixed-length sequences of adjacent textual items in the context of root-to-leaf paths. Feature selection without any tunable threshold is used to choose a subset of the XML features on the basis of their relevance to clustering, which is assessed through a new scoring scheme. A comparative experimentation on real-world benchmark XML corpora reveals a higher effectiveness than several state-of-the-art competitors.

We propose a generative probabilistic approach to modeling interactions in directed graphs for unveiling the participation of nodes in multiple communities along with the roles played therein. Precisely, a hierarchical Bayesian model is developed, in which node interactions are governed by latent explicative reasons regarded as personal and contextual interaction factors. The former are inherently descriptive of nodes, while the latter are characterizations of individual communities and roles. The generative process of the devised model assigns nodes to communities with respective roles and connects them through directed links, that are ruled by their personal and contextual interaction factors.We derive posterior inference for the presented model based on approximate Markov-Chain Monte-Carlo methods. Finally, we demonstrate via a comparative analysis on real-world networks the superior performance of our approach in terms of community compactness and predictive power of the discovered communities, roles and interaction factors.

XML co-clustering is a promising method to overcome the effectiveness of traditional XML clustering approaches, due to the exploitation of the mutual relationships between XML documents and their respective XML features while clustering both simultaneously. To shed light on this so far unexplored research direction, we conduct a systematic study of the effectiveness of XML co-clustering, by viewing the task as parametric with respect to the XML features. Thus, the definition and exploitation of three distinct types of XML features, which are respectively informative of the content, structure and both aspects of the XML documents, allows an in-depth investigation of all three different instances of the XML co-clustering task, i.e., XML co-clustering by content alone, structure alone as well as both structure and content. XML co-clustering relies on a non-negative matrix trifactorization technique, that efficiently processes large-scale input data, which is especially useful with large corpora of text-centric XML documents. The relevance of the structural and content features of the XML documents is assessed through a new weighting scheme. An intensive experimental evaluation on real-world benchmark XML corpora reveals a higher effectiveness of XML co-clustering in comparison with state-of-the-art approaches to XML clustering. Insights are also provided on the effectiveness of XML feature clustering.

Non-negative matrix factorization is intensively used in text clustering. We investigate its exploitation in the XML domain for clustering XML documents by structure and content into topically homogeneous groups. Non-negative matrix factorization is performed through an alternating least squares method, which incorporates expedients to attenuate the burden of large-scale factorizations. This is especially relevant when massive text-centric XML corpora are processed. Empirical evidence from a comparative evaluation on real-world XML corpora reveals that our approach overcomes several state-of-the-art competitors in effectiveness.

The combination of multiple clusterings for partitioning XML documents is proposed as a promising method, aimed to decompose the inherently difficult problem of catching structural and content relationships within an XML corpus into a number of simpler subproblems. To verify the validity of such an intuition, a new technique for partitioning XML documents is presented, in which conventional clustering techniques operating on flattened representations of individual aspects of the XML documents (that also include some rare patterns) are used to partition the available XML corpus. The effectiveness of the devised technique is revealed by a comparative empirical evaluation on benchmark XML corpora.

The supervised classification of XML documents by structure involves learning predictive models in which certain structural regularities discriminate the individual document classes. Hitherto, research has focused on the adoption of prespecified substructures. This is detrimental for classification effectiveness, since the a priori chosen substructures may not accord with the structural properties of the XML documents. Therein, an unexplored question is how to choose the type of structural regularity that best adapts to the structures of the available XML documents. We tackle this problem through X-Class, an approach that handles all types of tree-like substructures and allows for choosing the most discriminatory one. Algorithms are designed to learn compact rule-based classifiers in which the chosen substructures discriminate the classes of XML documents. X-Class is studied across various domains and types of substructures. Its classification performance is compared against several rule-based and SVM-based competitors. Empirical evidence reveals that the classifiers induced by X-Class are compact, scalable, and at least as effective as the established competitors. In particular, certain substructures allow the induction of very compact classifiers that generally outperform the rule-based competitors in terms of effectiveness over all chosen corpora of XML data. Furthermore, such classifiers are substantially as effective as the SVM-based competitor, with the additional advantage of a high-degree of interpretability.

We present a new probabilistic approach to modeling social interactions, that seamlessly integrates community discovery and role assignment for a deeper understanding of connectivity patterns in social networks. The devised approach is an unsupervised learning technique based on a Bayesian hierarchical model of social interactions. This model specifies an intuitive generative process, in which pairs of nodes in a social network are associated with communities as well as roles in the context of the respective communities, before that a directed interaction is possibly established between them. According to the generative semantics of the proposed model, nodes are represented as probability distributions over communities, while communities are represented as probability distributions over roles. Such distributions are unknown parameters of the proposed model, that are estimated from social-network data through approximated posterior inference and parameter estimation. A comparative evaluation over real-world social networks reveals that our approach outperforms state-of-the-art competitors in terms of link prediction.

XML clustering by structure is, in its most general form, the process of partitioning a corpus of XML documents into disjoint clusters, such that intra-cluster structural homogeneity is high and inter-cluster structural homogeneity is low. In this paper, we propose an algorithm that implements a partitioning strategy, in which root-to-leaf paths are used to separate the XML documents. Paths are discriminatory substructures and, thus, the effectiveness of our algorithm is accordingly high. Moreover, a suitable encoding is adopted for representing and testing the occurrence of the individual paths within each XML document independently of the length of such paths. Not only this expedites clustering, but it also makes our algorithm scalable to process large-scale corpora of XML documents. A comparative evaluation over several standard (real-word and synthetic) XML corpora reveals that our algorithm outperforms all of its competitors in efficiency and scalability, while being as effective as the top-notch competitors. One especially appealing property of the proposed algorithm is that it achieves these levels of performance by automatically establishing a natural number of clusters to be discovered in the underlying XML corpus.

We propose a new approach to XML classification, that uses a particular rule-learning technique for the induction of interpretable classification models. These separate the individual classes of XML documents by looking at the presence within the XML documents themselves of certain features, that provide information on their content and structure. The devised approach induces classifiers with outperforming effectiveness in comparison to several established competitors.

We propose a bayesian probabilistic model for explicit preference data. The model introduces a generative process, which takes into account both item selection and rating emission to gather into communities those users who experience the same items and tend to adopt the same rating pattern. Each user is modeled as a random mixture of topics, where each topic is characterized by a distribution modeling the popularity of items within the respective user-community and by a distribution over preference values for those items. The proposed model can be associated with a novel item-relevance ranking criterion, which is based both on item popularity and user's preferences. We show that the proposed model, equipped with the new ranking criterion, outperforms state-of-art approaches in terms of accuracy of the recommendation list provided to users on standard benchmark datasets.

A methodology for clustering multi-relational data is proposed. Initially, tuple linkages in the database schema of the multi-relational entities are leveraged to virtually organize the available relational data into as many transactions, i.e. sets of feature-value pairs. The identified transactions are then partitioned into homogeneous groups. Each discovered cluster is equipped with a representative, that provides an explanation of the corresponding group of transactions, in terms of those feature-value pairs that are most likely to appear in a transaction belonging to that particular group. Outlier data are placed into a trash cluster, that is finally partitioned to mitigate the dissimilarity between the trash cluster and the previously generated clusters.

In this work we propose DAEDALUS, a formal framework and system, specifically focussed on progressive combination of mining and querying operators. The core component of DAEDALUS is the MO-DMQL query language that extends SQL in two respects, namely a pattern definition operator and the capability to uniform manipulating both raw data and unveiled patterns. DAEDALUS system is specifically focussed on movement data and has been implemented as a query execution layer on top of the Hermes Moving Object Database. The expressiveness and usefulness of the MODMQL language as well as the computational capabilities of DAEDALUS are qualitatively evaluated by means of a case study.