Search

In this study, we propose the User2Vec framework, a novel approach for capturing user behavior based on the co-occurrence of user activities. User behavior modeling is essential for understanding relationships in areas such as e-learning and marketing, but most existing methods rely on explicit social interaction data, which are not always available. We propose the User2Vec framework, inspired by the Word2Vec model, to encode user activity into dense vector representations using the temporal co-occurrence of active users. This approach allows the identification of user groups without relying on direct interaction data. The proposed methodology involves preprocessing user activity logs, extracting temporal co-occurrence features, and applying clustering techniques such as k-means and spectral clustering to reveal latent structures. These clustering methods were chosen for their balance of computational efficiency and effectiveness in capturing implicit relationships. The approach is experimented on a dataset from a university e-learning platform, where student logs are processed to cluster users according to their co-occurring actions. Comparative evaluations show that the proposed model outperforms previous co-occurrence-based methods in terms of clustering accuracy and computational efficiency. Our results highlight the significant potential of the proposed framework in applications such as personalized learning and community detection.

This paper presents a model for the automatic classification of writing proficiency in Italian as a second language (L2) according to the Common European Framework of Reference for languages (CEFR). The proposed method integrates lexical and morphosyntactic quantitative analysis with phraseological dimensions. Phraseological aspects include the ability to use and understand fixed expressions, idioms, and other multi-word units that are common in a language and reflect the depth of language comprehension typically manifested by native speakers. Specific techniques for encoding phraseological features have been introduced, and basic phraseological statistics, previously unavailable for Italy, have been extracted from an Italian corpus. The proposed model was experimentally compared with widely used machine learning models on a dataset of written texts produced by non-native speakers for the official Italian CEFR certification exams. The experimental results outperformed previous work on the CEFR classification of Italian L2 proficiency in terms of accuracy and all relevant prediction metrics, demonstrating the effectiveness of the proposed approach, which integrates morphosyntactic and phraseological features.

This study addresses the problem of how to automatically generate source code that is not only functional, but also well-structured, readable, and maintainable. Existing generative models for source code often produce functional code, but they lack consistency in structure and adherence to coding standards, essential for integration into existing application development projects and long-term software maintenance. By training the model on specific code structures, including a dataset with Italian annotations, the proposed methodology ensures that the generated code is compliant with both the functional requirements and the pre-defined coding standards. The methodology proposed in this study applies transfer learning techniques on the DeepSeek Coder model, to refine pre-trained models to generate code that integrates additional structuring constraints. By training the model on specific code structures, including a dataset with Italian comments, the proposed methodology ensures that the generated code meets both functional requirements and coding structure. Experimental results, evaluated using the perplexity metric, demonstrate the effectiveness of the proposed approach, which impacts the goals of reducing errors, and ultimately improves software development quality.

The future of education lies in the ability to develop technologies which integrate seamless artificial intelligence (AI) components into the educational process, in order to deliver a personalized service which is dynamically tailored to the learner’s characteristics, abilities, and needs [...]

Deep learning approaches for facial Emotion Recognition (ER) obtain high accuracy on basic models, e.g., Ekman’s models, in the specific domain of facial emotional expressions. Thus, facial tracking of users’ emotions could be easily used against the right to privacy or for manipulative purposes. As recent studies have shown that deep learning models are susceptible to adversarial examples (images intentionally modified to fool a machine learning classifier) we propose to use them to preserve users’ privacy against ER. In this paper, we present a technique for generating Emotion Adversarial Attacks (EAAs). EAAs are performed applying well-known image filters inspired from Instagram, and a multi-objective evolutionary algorithm is used to determine the per-image best filters attacking combination. Experimental results on the well-known AffectNet dataset of facial expressions show that our approach successfully attacks emotion classifiers to protect user privacy. On the other hand, the quality of the images from the human perception point of view is maintained. Several experiments with different sequences of filters are run and show that the Attack Success Rate is very high, above 90% for every test.

Climate change threats make it difficult to perform reliable and quick predictions on floods forecasting. This gives rise to the need of having advanced methods, e.g., computational intelligence tools, to improve upon the results from flooding events simulations and, in turn, design best practices for riverbed maintenance. In this context, being able to accurately estimate the roughness coefficient, also known as Manning’s n coefficient, plays an important role when computational models are employed. In this piece of research, we propose an optimal approach for the estimation of ‘n’. First, an objective function is designed for measuring the quality of ‘candidate’ Manning’s coefficients relative to specif cross-sections of a river. Second, such function is optimised to return coefficients having the highest quality as possible. Five well-known meta-heuristic algorithms are employed to achieve this goal, these being a classic Evolution Strategy, a Differential Evolution algorithm, the popular Covariance Matrix Adaptation Evolution Strategy, a classic Particle Swarm Optimisation and a Bayesian Optimisation framework. We report results on two real-world case studies based on the Italian rivers ‘Paglia’ and ‘Aniene’. A comparative analysis between the employed optimisation algorithms is performed and discussed both empirically and statistically. From the hydrodynamic point of view, the experimental results are satisfactory and produced within significantly less computational time in comparison to classic methods. This shows the suitability of the proposed approach for optimal estimation of the roughness coefficient and, in turn, for designing optimised hydrological models.

Evolutionary algorithms and metaheuristics are widely used to provide efficient and effective approximate solutions to computationally difficult optimization problems [...]

This work proposes an innovative visual tool for real-time continuous learners analytics. The purpose of the work is to improve the design, functionality, and usability of learning management systems to monitor user activity to allow educators to make informed decisions on e-learning design, usually limited to dashboards graphs, tables, and low-usability user logs. The standard visualisation is currently scarce, and often inadequate to inform educators about the design quality and students engagement on their learning objects. The same low usability can be found in learning analytics tools, which mostly focus on post-course analysis, demanding specific skills to be effectively used, e.g., for statistical analysis and database queries. We propose a tool for student analytics embedded in a Learning Management System, based on the innovative visual metaphor of interface morphing. Artificial intelligence provides in remote learning immediate feedback, crucial in a face-to-face setting, highlighting the students’ engagement in each single learning object. A visual metaphor is the representation of a person, group, learning object, or concept through a visual image that suggests a particular association or point of similarity. The basic idea is that elements of the application interface, e.g., learning objects’ icons and student avatars, can be modified in colour and dimension to reflect key performance indicators of learner’s activities. The goal is to provide high-affordance information on the student engagement and usage of learning objects, where aggregation functions on subsets of users allow a dynamic evaluation of cohorts with different granularity. The proposed visual metaphors (i.e., thermometer bar, dimensional morphing, and tag cloud morphing) have been implemented and experimented within academic-level courses. Experimental results have been evaluated with a comparative analysis of user logs and a subjective usability survey, which show that the tool obtains quantitative, measurable effectiveness and the qualitative appreciation of educators. Among metaphors, the highest success is obtained by Dimensional morphing and Tag cloud transformation.

In this paper, a Neural Networks optimizer based on Self-adaptive Differential Evolution is presented. This optimizer applies mutation and crossover operators in a new way, taking into account the structure of the network according to a per layer strategy. Moreover, a new crossover called interm is proposed, and a new self-adaptive version of DE called MAB-ShaDE is suggested to reduce the number of parameters. The framework has been tested on some well-known classification problems and a comparative study on the various combinations of self-adaptive methods, mutation, and crossover operators available in literature is performed. Experimental results show that DENN reaches good performances in terms of accuracy, better than or at least comparable with those obtained by backpropagation.

This article presents a novel hybrid classification paradigm for medical diagnoses and prognoses prediction. The core mechanism of the proposed method relies on a centroid classification algorithm whose logic is exploited to formulate the classification task as a real-valued optimisation problem. A novel metaheuristic combining the algorithmic structure of Swarm Intelligence optimisers with the probabilistic search models of Estimation of Distribution Algorithms is designed to optimise such a problem, thus leading to high-accuracy predictions. This method is tested over 11 medical datasets and compared against 14 cherry-picked classification algorithms. Results show that the proposed approach is competitive and superior to the state-of-the-art on several occasions.